1245 Commits
Author SHA1 Message Date
Willie 330114bd6a fix(install): use frontmatter names for generic skill directories (#617)
* fix(install): use frontmatter names for generic skill directories

Resolve meaningful defaults for generic containers while preserving explicit names, tracked paths, existing update destinations, and directory-based source selection. Fall back on unsafe or colliding names to prevent overwrites.

Cover global and project installs, local and Git sources, naming overrides, tracked reports, collision safety, and updates. Closes #616.

* fix(install): preserve names during source-based updates

Resolve source-based updates against existing metadata before choosing destinations. Reject ambiguous or colliding update targets before writes, and reserve case-insensitive discovery names across filesystems.

Add global/project regressions for local and Git source updates, legacy destinations, grouping, and upstream-added collisions.

* fix(install): retain update ownership and selection checks

Only allow a metadata-verified same-source update to bypass destination conflicts. Reject occupied fallback destinations for both local and Git sources unless overwrite is explicitly forced.

Resolve destinations after skill selection so unrelated conflicts do not block filtered updates. Cover cross-source and unmanaged destinations, selected/excluded updates, and collision aborts in global and project modes.
2026-10-23 20:04:45 +08:00
Rusty Shackleford d23a44aec9 refactor(projects): share safe dashboard and CLI operations (#602) 2026-10-23 02:07:54 +08:00
Krishnaswamy SubramanianandWillie dabe06a297 feat(targets): support OpenCode config directory accounts (#594)
* feat(targets): support OpenCode config directory accounts

OpenCode v2 introduces support for custom configuration directories
via OPENCODE_CONFIG_DIR, storing opencode.json and agent directories.

Previously, skillshare rejected OpenCode accounts configured with a
custom config directory because OpenCode was missing config_dir and
was excluded from accountAgents in hooks and MCP services.

- Define default config_dir (~/.config/opencode) in targets.yaml
- Add OpenCode to account agents and register OPENCODE_CONFIG_DIR
  across hooks and MCP service handlers
- Route OpenCode account MCP configuration to opencode.json inside
  the account config directory

Resolves #593.

* fix(targets): add OpenCode to schemas, completions and docs

Add opencode to the targetConfig.agent enum and config_dir description
in schemas/config.schema.json so configurations containing OpenCode
account targets pass schema validation.

Include opencode in the target --agent completion scripts for Bash,
Zsh, Fish, PowerShell, and Nushell.

Update website documentation and built-in skill references to describe
OpenCode accounts and OPENCODE_CONFIG_DIR management alongside other
account agents.

* fix(targets): support plugin accounts and overlap

Add OpenCode account recognition and OPENCODE_CONFIG_DIR resolution to
the plugin service so plugin sync works for OpenCode account targets.

Introduce TargetScanPaths to resolve cross-runtime scan paths using
TargetConfig.Agent for account targets while retaining the account's
own primary directory. Update path overlap detection, doctor checks,
and LeakedSkills to warn about leaked skills for account targets.

Update localized docs (zh-Hans, zh-Hant, ja, ko) and plugin references
to document OpenCode account targets and OPENCODE_CONFIG_DIR. Update
the available targets handler test to expect OpenCode's configDir.

* fix(targets): isolate OpenCode accounts and scan paths

Reject OPENCODE_CONFIG overrides when operating on OpenCode accounts
and clear OPENCODE_CONFIG in the account execution environment so
plugin operations target the account directory.

In TargetScanPaths, remap config-root paths to the account directory
while preserving external scan paths like Codex's ~/.agents/skills.

* fix(plugins): use account target in OpenCode update probe

Pass c.Target instead of a hardcoded "opencode" when probing whether
OpenCode supports plugin updates, ensuring account-configured CLI
binaries and account environment variables are used during preview.

* test(mcp): isolate XDG_CONFIG_HOME in account test service

Isolate XDG_CONFIG_HOME to the temporary test directory in
agentAccountService to prevent runner environment variables from leaking
into TestAccountEnvOverrideDoesNotPruneDefaultHome when testing OpenCode
account shell overrides.

---------

Co-authored-by: Willie <egg358358@gmail.com>
2026-10-22 21:50:06 +08:00
Willie 2bba224cdf fix(ui): update agents from the Updates tab and repo detail page (#601)
The Agents Updates tab updates through /api/update/stream, which resolved
every name against the skills source, so agent updates failed there. The
agent detail page names a tracked repo by its folder, which updateAgent
did not resolve, so updating an agent inside a repo failed with "agent not
found".

The stream now takes a kind and routes agent names to updateAgent, which
also resolves an agent repo folder under the agents source. The tab sends
its kind to the stream and to Prune, and matches update results by kind and
by the agent's path, so a nested agent is marked up to date after it
updates.

Closes #600
2026-10-22 00:39:12 +08:00
Willie 7b952c02a9 fix(update): honor accepted findings in dashboard tracked updates (#599)
* fix(update): honor accepted findings in dashboard tracked updates

The dashboard ignored findings accepted through the CLI, so a tracked repo
that `update --force` let through was blocked again on the next dashboard
update, and every dashboard Force Retry had to be repeated. Regular skill
updates in the dashboard already honored and recorded acceptances, so only
tracked repos differed.

Tracked-repo updates now always honor accepted findings and record the ones
an override lets through, for both callers. The opt-in option and the
AuditGate flag it fed are removed, since no caller turns them off.

Dashboard agent tracked repos passed an empty source dir, so a nested repo
kept acceptances in its group folder under a different key than the CLI.
They now pass the agents source, as the CLI does.

* fix(update): reload the agents store after agent repo refresh

Dashboard agent tracked repos now refresh metadata in the agents source,
but a metadata change reloaded the skills store, leaving the agents store
as it was until the next request. Reload the store that matches the source.

Every caller of update.TrackedRepo now passes a source dir, so the
empty-dir guard and its doc no longer describe a supported mode. The new
agent test reuses the tracked fixture instead of rebuilding it.
2026-10-21 23:28:16 +08:00
star-nebula e10304eb81 test(theme): enable Windows CI coverage (#589)
Correct Unix-only resolution expectations on Windows and pin terminal and renderer state so the package is deterministic. Add theme to the Windows test job to catch future platform regressions. Related to #344.
2026-10-21 23:08:20 +08:00
Willie 829cbb176d fix(plugins): honor native OMP marketplace builds (#588)
* fix(plugins): honor native OMP marketplace builds

Resolve OMP entries from their native marketplace when present and inspect package.json extension-only plugins. Keep target-specific source paths, marketplace identities and components so Claude adoption and hybrid manifests remain valid.\n\nShow separate builds in text discovery without changing logical JSON selection, and cover native precedence, invalid catalogs, manifestless and hybrid packages, global/project materialization and CLI output with regression tests. Update the plugin docs and existing translations.\n\nFixes #587

* fix(plugins): preserve native build metadata across consumers

Keep OMP catalog versions in target discovery metadata and materialized marketplace entries so previews, cache guards, and native installs use the same version, including legacy catalog fallback. Cover catalog versions that differ from package manifests.

Show each dashboard discovery target's own version, components, and source path instead of presenting a merged candidate as one build.

* fix(plugins): avoid ambiguous shared build versions

Derive candidate versions from agreement across supported targets and recompute after catalog overrides. No-Agent records and lists no longer inherit an arbitrary native build version; equal versions remain available and unknown versions stay unknown.

Keep Cursor and Antigravity inventory target-specific when manifests disagree. Cover no-Agent preview and persistence, equal and unknown versions, catalog overrides, and mixed-manifest local inventories. Align English and translated plugin references.

* fix(plugins): accept Pi extension manifests for OMP catalogs

Reuse the existing OMP extension detector instead of requiring an OMP-specific manifest key. Native OMP accepts Pi extension manifests too, so the extra gate rejected supported native catalog entries.

Cover Pi-only discovery with scoped package names and global/project materialization. Align the English plugin references and all existing translations.
2026-10-20 12:22:05 +08:00
Willie f6e7265414 feat(hooks): read hook code and scripts from files next to the config (#586)
* feat(hooks): read hook code and scripts from files next to the config

Code Agent extensions and command scripts lived as strings in
config.yaml, so a longer Pi extension or shell script was edited inside
a YAML block scalar with no editor support, and nothing type-checked or
linted it. A breaking change in an Agent's extension API went unnoticed
because the generated file is not part of any tsconfig.

A binding now takes `codeFile: path` instead of `code`, and a `files`
value may be `{from: path}` instead of the contents. Paths resolve
against the folder of the config that holds the entry and are read
every time the config is read, so the regular typecheck or shellcheck
can cover the real source. The revision and drift checks use the
resolved contents, so a file edited after a preview fails the apply as
stale.

Saving keeps the references: enable/disable, `hooks edit` and the
dashboard write `codeFile` and `{from}` back instead of inlining the
contents. JSON carries the contents plus `codeFile`/`filesFrom`, and an
entry echoed back with both must match the file, so an edit is refused
rather than silently dropped. Paths must stay inside the config folder
without symlinks, so a cloned project cannot pull in files from
elsewhere. References are taken from the YAML tree before one strict
decode, which keeps unknown-field checks and anchors shared across
bindings working.

Closes #585

* fix(hooks): refuse an emptied file sent with its reference

An empty files value stood for "contents not sent", so a file cleared in
the dashboard's Git YAML editor while its filesFrom reference stayed was
treated as unsent: the file was read again from disk and the save
reported success with the user's edit discarded.

A value read from {from: path} now keeps a dedicated mark until it is
resolved, so an empty value is compared with the file like any other
contents and refused when they differ. Code keeps using empty as unsent,
since a code binding can never be empty.

* fix(hooks): render referenced files as they are on disk

RenderNative rendered the contents the caller sent. The dashboard sends
the entry it loaded, so after a referenced file changed, View
configuration still showed the old contents while describing them as
what sync would write.

Render now drops sent contents for codeFile and {from} references and
reads every reference again, like preview and save. Draft, render and
config decoding are the only paths that take an entry, and all three now
resolve references.

* fix(hooks): accept file references in JSON entries and anchored values

Two inputs that config.yaml accepts failed to load elsewhere.

A `{from}` value carrying its own YAML anchor, aliased from another
binding, lost its anchor because the value node was replaced instead of
rewritten, so the alias no longer resolved. The node is now rewritten in
place with its anchor kept.

An entry JSON file (`hooks add|edit --file x.json`) could not write a
files value as `{"from": path}`, because JSON decoded straight into
string contents. Binding now decodes JSON files values as contents or a
reference, strictly, like YAML, so both entry formats and the Web API
take the same shapes.
2026-10-19 22:48:00 +08:00
Willie 36d04b526b fix(hooks): keep a YAML block scalar's final newline in entry files (#584)
ParseEntry trimmed the whole document before decoding it, which was
meant only to detect a leading `{` for JSON. The YAML branch decoded
the trimmed bytes, so a `code: |` block scalar ending the file lost
its final newline: under clip chomping the line break survives only
if it is present. A Pi or OMP extension added from such a file was
written without a trailing newline and failed formatters such as
`prettier --check`. JSON input was unaffected because its newline is
escaped inside the string.

The YAML branch now decodes the original bytes.
2026-10-19 15:42:16 +08:00
Willie 70ceacca9b feat(mcp): serve an Executor skill index from mcp serve --http (#583)
* feat(mcp): serve an Executor skill index from mcp serve --http

Executor apps load remote skills with wellKnownSkills, but could only
reach Skillshare's skills through a copy in Git, so an edit saved with
apply_skill_edit never reached them without a commit and push. With
--http, mcp serve now also answers /.well-known/agent-skills/index.json
and the files it lists, built from the same catalog and token, so
Executor sees an edit on its next check.

The format is the one Executor's loader reads, not the agent-skills
discovery draft v0.2. One skill that loader rejects fails its whole
load, so the index leaves out skills it would reject (non-text metadata,
license or allowed-tools, files over 2 MB, non-UTF-8 files, file names
decodeURIComponent cannot handle, names used twice) and lists them in
--check and on stderr at startup. Versions are digests of each skill's
paths and file digests, and the output is stable while nothing changes,
because the loader reads the index twice and compares the bytes. An
index over Executor's file or byte limits answers with an error naming
--target.

Manifest files now record their path within the listing skill: a
parent's entry otherwise took its nested skill's paths, which would
have listed SKILL.md twice. The index route repeats the MCP handler's
DNS rebinding check for loopback servers without a token.

Closes #574

* fix(mcp): accept proxied hosts with a token and use Executor's app accounts

Behind the documented TLS proxy, a loopback mcp serve sees the proxy's
public Host, and the DNS rebinding check answered 403 even with a valid
bearer token. The index route had it, and so did the MCP endpoint through
the go-sdk's localhost protection. A rebinding page cannot send the token,
so both checks now apply only when no token is set; the docs say a
tokenless server behind a proxy needs a loopback Host.

The Executor example used the account form apps 0.0.1-beta.90 removed
(`{ accounts }` and `ctx.accounts`, now typed as never). It declares the
token as an app account and reads ctx.app.accounts, in all five locales.
The wellKnownSkills loader itself is unchanged at the v2 head (b98424d9).
2026-10-18 23:48:42 +08:00
Willie 578790c416 feat(mcp): write per-Agent native fields from the source (#579)
* feat(mcp): write per-Agent native fields from the source

Some MCP settings have no portable form, such as Codex's
default_tools_approval_mode or Claude Code's per-server timeout. Sync
kept them only while an entry survived, so a recreated entry (for
example after a rename) lost them, and every machine needed them added
by hand again.

A server's `options` now holds these fields per Agent, and sync writes
them as given. The field ledger that piOptions used is extended to
every Agent, so sync removes an option it wrote once the source drops
it, and reports a conflict when the Agent changed it since.

Options are refused when they could break the entry or leak a secret:
fields Skillshare writes (including portable names some Agents rename),
values a native format cannot write back (null, arrays of tables),
options.pi (piOptions covers Pi), disabled entries, and credentials as
plain text, since the Agent's file may be committed.

The dashboard has no editor for options yet but keeps them on save.

Refs: #578

* fix(mcp): close gaps in per-Agent options found in review

- A credential option passed if it merely contained "${", so text around
  a reference, or another Agent's syntax, was written out as plain text.
  It must now be a whole-value reference in the Agent's own syntax, and
  Codex and Antigravity options, which read no references, cannot hold one.
- The Kilo Code project check looked only at portable env references, so
  an {env:VAR} in options reached a project file Kilo refuses to load.
- A replace resolution left the dropped-option conflict in place, so the
  plan stayed blocked; replace now lets the source win and prunes those
  fields.
- Adopting an entry that already matched skipped writing its options until
  a later sync; the takeover now writes them.

* fix(mcp): refuse numeric credentials and constrain option values in the schema

- YAML reads an unquoted secret such as clientSecret: 123456 as a number,
  which the credential check skipped, so sync wrote it out as plain text.
  A mapping under a sensitive name such as credentials also hid its
  plainly named fields. Sensitivity now carries down through mappings and
  lists, and a number under a sensitive name is refused.
- The schema allowed any value inside an Agent's options, so editors
  accepted null or nested lists that LoadSource then rejects. A recursive
  optionValue definition now matches the runtime rules.

* docs(mcp): name every Agent whose options cannot hold a credential

The docs named only Codex and Antigravity, but Claude Desktop, Goose,
Junie, LM Studio and Warp read no environment references there either,
so a source such as options.goose.apiKey: "${GOOSE_KEY}" looked valid
while LoadSource refused it.
2026-10-17 23:30:04 +08:00
Willie 4a5dacac6b feat(mcp): let mcp serve clients edit listed skills (#577)
* feat(mcp): let mcp serve clients edit listed skills

A remote assistant that reaches skills only through `mcp serve` could
read them but had to hand every change back to a person (#540). Skills
listed in `mcp.serve.edit.skills` can now be changed one existing file
at a time: `preview_skill_edit` writes nothing and returns the full
diff, the audit result and a preview id, and `apply_skill_edit` saves
exactly that preview, once.

Every save is tied to the revision the client read, so a file changed
in the meantime fails with a conflict instead of being overwritten.
Skills that `update` would overwrite stay read-only: tracked repos,
upstream installs, and skills nested inside an install. The skill is
validated and audited in a temporary copy before anything is written;
as with install, a finding at or above `audit.block_threshold` refuses
the edit unless it was accepted with `--force`, and findings are marked
new or existing so the client can see what its change adds.

The write is atomic, keeps the file mode, goes through sourcefs link
checks, and is recorded in the operation log. Native Skills extension
clients still see the edit tools, since the extension cannot write.

* fix(mcp): close gaps in mcp serve skill editing

Codex review of #577 found five ways the edit tools could do more, or
record less, than they promise:

- A blank pattern in mcp.serve.edit.skills was dropped by the filter,
  and no patterns match every skill, so `skills: [' ']` made every skill
  editable. Blank patterns are now a config error.
- preview_ttl was only checked when skills were listed, so the dashboard
  could save an invalid value that later stopped mcp serve. It is now
  checked whenever edit is present.
- Apply compared only the edited file's digest, so two previews of
  different files, or an edit made elsewhere in between, could save a
  skill state that was never validated or audited. Apply now checks and
  audits the skill again as it will be saved.
- Edit audits were not written to audit.log, so blocked attempts left no
  trace in `skillshare log --audit`. Each preview and apply audit is now
  logged like install's audit gate.
- The writability check migrated legacy metadata, so a "writes nothing"
  preview (and --check) could rewrite .metadata.json and delete
  registry.yaml. It now reads the store without migrating; a legacy
  registry.yaml keeps every skill read-only, and a legacy sidecar marks
  its skill as installed.
2026-10-15 22:30:39 +08:00
Willie fdca87a1ff feat(mcp): mark the mcp serve skill tools read-only
Clients and gateways that gate tools by annotation treat a tool without
readOnlyHint as one that changes data. Executor v2 does, so every
list_skills or read_skill call could stop for approval even though both
only read the skills source. Marking them read-only lets those clients
call them without asking. Refs: #540.
2026-10-15 21:34:45 +08:00
Willie 0c69dad6c2 fix(update): stop trusting tree hashes for skills with symlinks (#573)
* fix(update): stop trusting tree hashes for skills with symlinks

Install copies a file symlink's target, but git stores the link as a blob
holding the target path, so a skill's subtree hash stays the same when a
target outside the folder changes. Batch update skipped such skills as
current and check reported them up to date, leaving stale content installed.

A subdir that holds a symlink now has no usable tree hash: install records
none, update falls back to a reinstall, and check reports the skill as
update_available. Check lists entries with -z so links with non-ASCII names
are not missed behind git's path quoting.

Closes #572

* fix(check): split ls-tree -z output on NUL only

parseLsTreeOutput also split on newlines so the old newline fixtures kept
passing, but with -z a newline belongs to the path. A directory or symlink
whose name holds a newline was cut in two, and a symlink under it was missed.
The fixtures now use NUL-terminated entries like real -z output.

Also writes both shortcut comments in the repo's one-line form and uses
testutil.RunGit in the new update test instead of a raw exec call.

* docs(update): note that skills with symlinks always reinstall

The update page says grouped updates leave skills whose folder did not
change alone. A skill with a symlink no longer has a usable tree hash, so it
is reinstalled whenever the upstream commit moves. State that exception on
the English page and its four translations.
2026-10-15 00:22:15 +08:00
Colin MasonandWillie c4b183f6ef fix(update): read store in batch fast path (#569)
* fix(update): read store in batch fast path

The grouped update fast path read per-skill .skillshare-meta.json
sidecars. The migration to .metadata.json deletes those files, so the
check never matched and update --all reinstalled every skill on every
run.

Read the central store instead, and refresh version and commit there
instead of writing a sidecar. Require SKILL.md on disk so the update
restores a deleted install. --force now skips the check and reinstalls
everything.

* test(update): cover legacy grouped metadata keys

Grouped skills can still be stored under a basename key with a Group
field. Check that the current-skill lookup finds that entry and that the
version refresh updates it in place instead of adding a new key.

* feat(update): report up-to-date skills

The summary lumped skills the fast path skipped in with failures. Count
them as "up to date" in the summary and as up_to_date in --json
output.

* feat(update): log up-to-date count in oplog

The operations log recorded updated, skipped, pruned, and blocked counts
for update but not the new up-to-date count. Record it as up_to_date
so the log matches the summary and --json output.

* test(update): cover up-to-date batch skills

update --all reinstalled every grouped skill on every run. Check that
unchanged skills report as up to date in global and project mode, that
an upstream change updates only the changed skill, and that --force
reinstalls everything.

* docs(update): describe up-to-date batch skills

Describe how update --all and --group leave unchanged skills alone and
report them as up to date, the up_to_date JSON field, and the new
meaning of --force.

* fix(update): keep lock pins on tree-only matches

A tree-hash match after upstream HEAD moved refreshed the full commit in
the store. A project lock pin follows that field, so an unrelated
upstream commit rewrote skills.lock.json even though the skill's files
did not change. Refresh only the version, as the old sidecar refresh
did. check compares the version, so it still reports the skill as up to
date.

The help text, docs and skill reference said --all and --group skip
current skills. The skip lives in the batch path, which runs for any
update of several skills, and a single skill always takes the reinstall
path. Say that instead.

* fix(update): load the store once per batch

Each skill skipped as current reloaded .metadata.json before comparing
its version, so an unchanged batch reparsed the whole store once per
skill. Compare the version already read from the snapshot first, and
refresh the skills that do lag HEAD with one load and save after the
loop. An unchanged batch of 400 skills drops from about 520ms to 64ms.

The help text and docs said any update of several skills takes this
path. In global mode, named and glob targets carry no metadata and
reinstall on the standalone path. Name what does take it: --all,
--group, and several skills named in project mode.

---------

Co-authored-by: Willie <egg358358@gmail.com>
2026-10-14 23:36:56 +08:00
Willie fe92a50b1b fix(upgrade): preserve UI assets for the pre-upgrade version (#571)
An already-running UI server still reads lazy-loaded chunks from its versioned cache after the CLI binary is upgraded. Preserve the pre-upgrade version alongside the downloaded version so upgrade cleanup does not break page navigation.

Add a multi-version cleanup regression test and document the retention policy and accepted consecutive-upgrade limitation across all website locales.
2026-10-14 23:10:44 +08:00
Willie 1094668408 fix(install): keep git tokens out of process arguments (#566)
Token auth rewrote the clone URL with url.<user>:<token>@host.insteadOf.
git hands the rewritten URL to git-remote-https as an argument, so any
user on the machine could read the token with ps for as long as a clone
or fetch ran. Every platform token (GitHub, GitLab, Bitbucket, Azure
DevOps, Gitea, CNB) went through the same rewrite.

git now gets the token from a credential helper scoped to the host,
which reads it from SKILLSHARE_GIT_AUTH_PASSWORD; printf is a shell
builtin, so the token never reaches an argv. An empty helper entry
first clears the user's own helpers for that host, so git neither asks
their keychain nor erases from it on a 401, and no longer stores the
skillshare token there as the URL rewrite did.

An http.extraHeader was tried first and dropped: the header goes out
before any 401, so a stale token broke installs from public repos, and
after a rejection git fell back to the user's helper and erased their
stored credentials.

Refs #565
2026-10-14 13:36:09 +08:00
Willie ac1d4b4387 fix(list): show every target sync placed a skill or agent in (#562)
* fix(list): show every target sync placed a skill or agent in

The Targets row of the list detail panel only looked for a symlink named
after the skill's flat name in the target's skills folder. It therefore
left out targets that sync fills in copy mode, targets linked whole in
symlink mode, targets using standard or prefixed naming, and every agent
target (agents were looked up in the skills folder). Targets with skills
turned off were still listed.

The row now asks sync's own name resolution and manifest which entry a
target holds for the skill, and checks agents in their resolved agents
folder by mode or extension output. Each target is resolved once per load
so moving through a large source under standard naming stays fast.

* fix(list): drop a skill's targets as soon as it is disabled in the TUI

Pressing t updated the row's disabled flag but not the discovered skills
that sync's name resolution reads, so the open list kept showing the
skill's targets while a reopened list, which rediscovers it as disabled,
showed none. The toggle now updates that flag too, so both agree: a
disabled skill lists no target, since the next sync removes its entries.
2026-10-14 01:47:30 +08:00
Willie 6cc790965a feat(mcp): pin tool definitions in mcp check --live and warn when they change (#560)
* feat(mcp): pin tool definitions in mcp check --live and warn when they change

An Agent reads a server's tool names and descriptions as instructions,
and a server can change them after the user reviewed it (a "rug pull").
The MCP spec 2026-07-28 local-server security guide notes the protocol
does nothing about this and advises clients that re-prompt on a changed
definition; most clients do not.

`mcp check --live` now records one SHA-256 per tool over name,
description and inputSchema the first time a server answers, in
mcp/tool-pins.json of the state directory, keyed by scope and server.
Later probes report a changed or new tool as a `tool-change` warning and
a removed one as info, until `--accept-tools` (only with --live) records
the current definitions. Without --live the pins are never read or
written. An unreadable pin file is never rewritten, since it may hold
other servers' pins. The dashboard words both warnings in every locale.

* fix(mcp): keep schema numbers exact, name the scope to accept in, and log acceptance

Codex review on #560:

- Tool definitions were decoded through float64 before hashing, so a
  schema bound changed from 9007199254740992 to 9007199254740993 hashed
  the same and raised no warning. Numbers now keep their literal text.
- The suggested accept command left out -g/-p; run from inside a project
  after a global check, it would accept the project's server instead.
  It now names the scope of the check that warned.
- Accepting changed tool definitions approves a possible rug pull, but
  left no record. --accept-tools now writes an operations log entry with
  the servers it accepted.

* fix(mcp): log a tool acceptance only when its baseline was saved

Codex review on #560: when comparing succeeded but writing tool-pins.json
failed (lock held, state directory unwritable), the report kept its
"accepted" finding, so the operations log listed the server as accepted
with status ok although its old pins remained. Such a server is now
logged under `failed` and the entry's status is error.

* fix(mcp): log first pins as accepted and report errors in the accept entry

Codex review on #560:

- A check that accepted one server but found an error on another still
  logged status ok, while the command itself exited 1. A server with an
  error is now listed under `failed` and the entry's status is error.
- --accept-tools on a server's first probe records its baseline, but the
  finding says "pinned", so the entry listed nothing as accepted. A first
  pin counts as accepted now.

* chore(release): request version 0.25.11

The MCP pin checks (#559, #560) and the backup staging fix (#558) ship
as 0.25.11 instead of the computed 0.26.0, per maintainer decision.

Release-As: 0.25.11
2026-10-13 23:22:57 +08:00
Willie 4382a8c8c8 feat(mcp): warn when a local server runs an unpinned package or image (#559)
* feat(mcp): warn when a local server runs an unpinned package or image

An entry such as `npx -y some-mcp` fetches whatever the registry serves
at every Agent start, so a compromised release reaches the machine on the
next launch without anyone noticing. The MCP local-server security guide
(spec 2026-07-28) names version pinning as a baseline and points out
that a container tag can move while a digest cannot.

`mcp check` now adds a `pin` finding from command and args alone, with no
network: npx, uvx and `pipx run` with no version or `latest`, and
`docker`/`podman run` with no tag or `latest`, are warnings; an image
pinned by tag but not by `@sha256:` digest is info. Value-taking run
flags are skipped so hardened lines such as `--cap-drop ALL` are not read
as the image. Warnings keep the exit code at 0, as before.

* fix(mcp): check every npx package, word pin warnings in the dashboard, document them in doctor

Codex review on #559:

- npx accepts --package more than once and fetches each one, but the pin
  check stopped at the first, so `npx --package=lodash@4.17.21
  --package=cowsay cowsay` passed although cowsay is unpinned. Every
  --package/-p value is now checked.
- The dashboard words env, command and dns findings in the viewer's
  language but showed the new pin warning in English. It now has its own
  sentence for an unversioned package and for an untagged image in every
  locale. The dashboard hides info findings, so the digest note needs none.
- doctor promotes check warnings, so an unpinned launcher now shows there;
  its docs list the pin check.

The mcp check runbook gains a step for the pin check.

* fix(mcp): check uvx --with packages and ignore environment markers in pins

Codex review on #559:

- uvx --with (-w) installs each named package beside the tool, so an
  unpinned one moves just like an unpinned tool, but the pin check
  skipped it as a flag value. Every --with requirement is now checked
  next to the --from or command package.
- In `--from 'mcp-server; python_version == "3.12"'` the == belongs to
  the environment marker, not to the package, yet it read as a pinned
  version. Only the requirement before ';' is inspected now.

* fix(mcp): treat only exact versions as pins, and pin the docs' own examples

Codex review on #559:

- npm resolves a dist-tag (`@next`) or a range (`@^1.2.3`) to whatever
  release matches at launch, yet any selector but `latest` passed as
  pinned. An npx package is pinned now only by an exact semver version,
  the rule Pi bindings already use.
- `==1.*` is a prefix match in Python version specifiers, so
  `uvx --from 'mcp-server==1.*'` can pick up a new 1.x release; it is no
  longer read as a pin.
- The Context7, Playwright and filesystem examples in the docs and the
  built-in skill launched unpinned packages, so copying them produced the
  new warning at once. They now name the current registry version.
2026-10-13 22:42:49 +08:00
Willie cf5783ee88 fix(backup): reclaim abandoned staging directories safely (#558)
* fix(backup): reclaim abandoned staging directories safely

Interrupted backup copies leave hidden .snapshot-* directories that retention cleanup never discovers. Hold a shared process lock during copies and use an exclusive cleanup lock to reclaim orphaned staging without deleting active work.

Share retention planning with non-mutating dry runs, report busy manual cleanup in the CLI and API, and align dashboard messages and translated docs. Cover process interruption and automatic cleanup with regression tests and an executable E2E runbook.

Closes #556

* fix(backup): address staging cleanup review findings

Keep cleanup reachable for staging-only dashboard backups and disclose abandoned incomplete directory deletion in every confirmation locale. Cover the empty-state confirmation with a regression test.

Read staging entries before a read-only lock probe so first-use dry runs cannot count a newly active writer. Refresh entries only under the acquired lock and classify both backup-root reads in the filesystem guard.

* fix(backup): align cleanup guides and doctor status

Disclose abandoned incomplete directory removal in backup help, completion descriptions, and the daily backup guide with all four translations. Add help and completion regressions so these consumers stay aligned with cleanup.

Keep the doctor Backups row visible when only the persistent lock remains after cleanup, preserving the existing JSON check and covering the lock-only state.

* fix(backup): complete cleanup documentation disclosure

Cleanup reclaims abandoned incomplete directories regardless of retention, so retention-only summaries understated the action. Align the backup synopsis, option table and previews, related workflow guides, built-in reference and all four website translations. Scope the retention defaults to completed snapshots.
2026-10-13 22:42:05 +08:00
Willie bfc402fe96 feat(mcp): open an Agent's MCP config folder, and project files, from the dashboard (#554)
* feat(mcp): open an Agent's MCP config folder, and project files, from the dashboard

A user who wants their own editor (Notepad++ on Windows) found setting
both PATH and $EDITOR too much just to open a config file, and could not
find any entry point for the MCP files of an mcp.projects root (#538).

Each Agent row on the MCP page now also has Open folder, which opens the
folder holding the file in the file manager, so any program can open it
from there. Before the Agent has written its file, it opens the closest
folder that exists. The project MCP tab gets the same Agents list for
that project's files. The open endpoint resolves project paths from the
config, as it does global ones, so a request still cannot name another
file, and folder and file opens keep the same-machine and headless
guards. The file manager launch is shared with the extensions folder.

* fix(mcp): keep a project's existing Agent files listed for remote viewers

Codex review: the project MCP tab built its Agents list from the
project's targets plus the files this viewer could open. A viewer on a
LAN address or behind a proxy can open none, so an existing file of an
Agent outside the project's targets dropped off the list, although the
docs say it shows with its path and only the open buttons go.

The list now returns which files exist apart from whether this viewer
may open them, and the page shows the open buttons only when local.

* fix(mcp): list the Agents a project's own servers write to in its file rail

Codex review: a server that exists only in one project can pick Agents
outside the project's targets, and sync writes their project files, but
the project tab listed only the project's targets and files that already
existed. Before the first sync such an Agent had no row, so its path
could not be copied and its folder could not be opened.

The list now also takes the targets of the project's own servers. Global
servers reach a project only through its targets, so they add nothing.
The rule lives in mcpView as projectFileTargets, with a unit test.
2026-10-13 15:00:39 +08:00
Willie 2d5849a0ef fix(config): keep YAML anchors and aliases when a command writes the config (#551)
* fix(config): keep YAML anchors and aliases when a command writes the config

Every writer re-encodes the in-memory struct, which has no anchors, so a
command such as `target a --add-exclude` expanded `&shared` / `*shared`
under targets and projects and the lists stopped following each other.
The comment pass added for #525 already pairs each old node with the one
that replaced it; relinkAliases uses the same pairs to put anchors back and
turn unedited alias values into aliases again.

An anchored node keeps its anchor and the aliases that still hold its old
value follow its new value, as the file declares. An alias that was edited
keeps its own list and leaves the anchored one alone. Output is sorted by
name, so the anchor moves to the first place the group appears: an alias
may not precede its anchor.

Fixes #546

* fix(config): keep YAML merge keys when a command writes the config

A command rewrites config.yaml from the Go struct, which holds the merged
keys of `<<: *base` and not the merge itself, so every merge key was written
out expanded and later edits of the anchored mapping no longer reached the
mappings that merged it. relinkAliases already restored anchors and aliases
from the old/new node pairing; relinkMerge now does the same for merge keys,
including `<<: [*a, *b]`, merges in nested mappings and in project-config
target items.

A key stays written out when the old mapping set it itself, or when its value
is no longer the one its merge source gives it. yaml.v3 merges are shallow, so
such a key replaces the merged value as a whole and what is written always
decodes to the config the command meant to save. A source that was removed, or
that holds a key the mapping no longer has, is not merged back.

Output order is not the old file's order, so an anchor that would follow its
first merge key moves onto it (`<<: &base`). The next write pairs that mapping
with the alias that still shows its value, so the file stays stable.

The runbook steps for anchors now write their own config: mdproof re-runs the
setup before each step, so step 7 and 8 never saw the config of step 6
(Codex review on #551). Steps for merge keys are added the same way.

* fix(config): let edits of an anchored mapping that add or remove a key reach its mergers

A command that added a key to an anchored mapping (the first `agents:`
block via `target base --agent-mode copy`) made relinkMerge refuse the
merge source: the merging mapping's pre-edit data lacked the key, which
was read as "the command removed it". The merge key was dropped and the
merger written out expanded, so later edits of the anchor no longer
reached it.

A source is now refused only when the merging mapping had a key before
the write and the command removed it. A key the source gained does not
count.

The symmetric case is fixed with it: a key removed from the anchored
mapping stayed written out under a merger that still held the old merged
value. An unedited key that the merger did not set itself now follows
the source that gave it, whatever that source now says (a changed value,
a removed key, or the next source of `<<: [*a, *b]`). A key the merger
set itself, or that the command edited, stays written out.

Tests cover add, remove and change of a key on the anchored mapping and
on the merging mapping, a key from either source of a merge list, nested
merges, projects, project-config target items and chained merges. Runbook
step 11 covers the real binary.

* fix(config): keep every comment when a write relinks merge keys and aliases

Rewriting a merge key replaced the original merge items and covered keys with
new nodes and carried over only the comments of the key and the list itself, so
a comment on an item of `<<:` followed by `- *a # note` was lost on any write,
and so was the trailing comment of a removed key when the mapping was left with
keys the merge supplies.

Comments now follow the entry they were on: a merge item's to the alias that
replaces it, those of an item that is no longer merged to the next item, and
those of a key the merge supplies to the next key still written, or in front of
the merge key when none is. The merge list keeps its flow or block style so item
comments stay on their lines.

Sweeping the rest of the relink paths for the same loss found more places where
yaml.v3 drops or moves a comment: the comment after an anchor on a block mapping
collided with its first key's comment, a line comment on a block collection or
on a key with an empty value is not written, and an anchored collection's head
comment moved on the next write. These are settled once on the written tree, and
the body of an inline merge source gets its comments paired too.
2026-10-13 11:14:06 +08:00
Willie 9e935b35fb fix(ui): accept a quoted editor path in $EDITOR (#553)
Open in editor split $VISUAL and $EDITOR on whitespace, so a Windows
path such as "C:\Program Files\Notepad++\notepad++.exe" broke at the
space and no editor was found. A user asked to open files in Notepad++,
which is not on PATH by default (#538).

Double or single quotes now keep a field together. Backslashes stay
literal, since Windows paths use them.
2026-10-13 10:41:27 +08:00
Willie 7ee08b913d feat(mcp): open an Agent's MCP config file from the dashboard (#552)
* feat(mcp): open an Agent's MCP config file from the dashboard

Users asked to jump from the MCP page straight into a client's config
file, such as opencode.json or .cursor/mcp.json, instead of copying the
path and opening it by hand (#538).

The open uses the same editor chain as a skill's Open in editor, so the
dashboard opens files one way. The endpoint takes a target, not a path,
and only opens a file the MCP page already lists. It answers only
requests from the machine running the dashboard, because the editor
starts on that machine. The list now reports which files exist, so the
button is hidden for a client detected before it has a file.

* fix(mcp): hide Open in editor where the server would refuse it

Codex review: the dashboard can be opened through a LAN IP, but the
open endpoint answers only loopback requests, so the button showed and
every click returned 403, while the docs promised it would open.

The list now reports `openable` instead of `existing`: the files that
exist, and only when the request comes from the dashboard's machine. A
viewer elsewhere does not see the button, and the docs say where it
appears.

* fix(mcp): don't treat a reverse proxy's viewers as local for Open in editor

Codex review: the documented Nginx and Caddy setups connect from
127.0.0.1, so the loopback socket check let any browser behind the proxy
see the button and launch editors on the server.

The browser now counts as local only when the socket and the Host it
used are both loopback and no proxy forwarding header is present. Both
documented proxies keep the browser's Host and add X-Real-IP or
X-Forwarded-For, so their viewers no longer qualify.
2026-10-13 10:40:27 +08:00
Willie 656437c2b5 fix(sync): limit project adoption writes to moved skills (#548)
* fix(sync): limit project adoption writes to moved skills

A stale install record could trigger a full project reconcile during sync, adding unrelated remote installs to the manifest, ignore rules and lockfile. Skip project writes when nothing was re-keyed, and update only moved declarations, ignore rules and existing pins while preserving full explicit reconcile behavior.

Cover the visible project directory sharing its global source with byte-identical no-move checks and a real move that preserves unrelated entries and audit acceptances.

* fix(sync): preserve existing move destination declarations

The adoption-only config rewrite could duplicate a destination already declared in the project manifest. Keep that destination declaration and remove the moved old entry, matching full reconcile behavior while preserving the destination fields.

Add a regression that fails with duplicate entries and passes with the existing destination branch retained. Re-run make check after the fix.
2026-10-13 10:37:41 +08:00
Willie ffdf9e57ab fix(plugin): stop evaluating Pi rules whose meaning depends on the Pi version (#549)
Pi 1.2.0 (earendil-works/pi#10765) runs normalizePath on resource-filter
patterns before matching, so `~`, `~/`, `file://` and, on Windows, `~\`
and Git Bash/MSYS/Cygwin/WSL drive paths now match files that older Pi
never matched. Skillshare supports Pi 0.99.2 and later without knowing
which one evaluates the rules, so it reported a definite selection that
could be wrong and let the write guard reason from it. Such rules now
evaluate as unknown, and the dashboard names the rule instead of a
generic message.

Exact rules also missed Pi's normalizeExactPattern in every version: a
leading ".\" is dropped on all platforms and backslashes become slashes
on Windows. Skillshare only dropped "./", so `-.\extensions\a.ts`
looked unmatched and produced a duplicate missing row.
2026-10-13 09:50:39 +08:00
Willie fd18cda655 fix(plugin): read plugin sources through their real path on macOS (#544)
Adding a plugin from a Git URL to Oh My Pi on macOS was always blocked
with "linked or non-directory OMP path component", which the dashboard
shows as "The reviewed OMP runtime destination must be verified empty".
The source is cloned into os.MkdirTemp, which lives under /var/folders,
and /var links to /private/var. The OMP install check reads the source's
package.json through ompSafePath, which rejects any linked component, so
Skillshare's own temp clone failed it. A local source under a linked
parent failed the same way.

acquireRef now hands callers the resolved root of the temp clone and of
a local directory, while the recorded source stays as given. Links inside
the source are still handled as before; cleanup still removes the
original temp directory.
2026-10-13 03:33:28 +08:00
Willie 12dd041288 fix(plugin): accept OMP 18.6.1 and later instead of exactly 18.6.1 (#543)
* fix(plugin): accept OMP 18.6.1 and later instead of exactly 18.6.1

The OMP adapter pinned the native version to exactly 18.6.1, so every
newer release was refused: the dashboard reported "Installation requires
OMP 18.6.1" for a user on 18.8.7, and removal and standalone extension
selection were blocked the same way.

The sources the adapter depends on (marketplace cache/manager/types,
plugin-cli output, the project anchor in discovery/helpers.ts, extension
loading and the pi-natives file lock) are unchanged from v18.6.1 through
v18.10.2, so the pin is now a minimum. The install, removal and selection
gates share one check built on the existing plainVersion helper;
unparsable, prerelease and older versions stay blocked.

Messages, all dashboard locales, the plugin reference docs and the
built-in skill now state 18.6.1+. A test keeps the English guidance text
in en.json equal to the Go strings the dashboard matches on.

* fix(docs): record the OMP 18.6.1+ boundary in the cache safety report

The safety report still said installation is allowed only for OMP 18.6.1
and that an unknown version blocks it, which no longer matches the gate.
It now states the minimum, which versions stay blocked, and the source
comparison and native 18.8.7 run that justify replacing the pin. The two
OMP runbooks no longer require exactly 18.6.1.
2026-10-13 02:50:17 +08:00
Willie f540daf601 fix(mcp): turn off a global server in OpenCode 2 projects (#541)
* fix(mcp): turn off a global server in OpenCode 2 projects

OpenCode 2 replaces a global server with the project entry of the same
name, and silently drops an entry without its type and command or url.
The project switch Skillshare wrote, {"enabled": false}, therefore left
the server on in OpenCode 2, while OpenCode 1 merged it by field.

Checked against OpenCode 2.0.27 and 1.19.0: OpenCode 2 still reads the
flat mcp.<name> shape and turns enabled: false into disabled: true, and
both read an entry with type and command or url. So no version detection
is needed. When the global config under mcp.projects defines the server,
the OpenCode switch now carries its type and executable, or its url
without the query, next to enabled: false. Args, env, headers and the
query stay out of a project file that may be committed, as the Pi switch
once did. A switch that already carries the endpoint keeps it when the
global server later leaves the config, unless it was edited since sync.

In project mode, or when the global config lacks the server, only the
bare switch can be written; the plan now says OpenCode 2 ignores it, in
the CLI and the dashboard.

Import also reads OpenCode 2's mcp.servers section, which its own
`mcp add` writes, instead of reporting it as one broken server.

Refs #538

* fix(mcp): keep secrets in a URL path out of the OpenCode switch

The OpenCode switch dropped only the query of the global server's URL.
Some MCP services put the key in the path instead, and an absolute
executable path names the user's home directory; both then reached a
project file that may be committed. A disabled server never starts, and
OpenCode 2 accepts any type with a command or url, so the switch now
carries only the executable's name or the URL's scheme and host.

Refs #538

* fix(mcp): skip OpenCode 2's timeout defaults when importing

OpenCode 2 keeps server defaults in mcp.timeout next to mcp.servers, so
import listed timeout as a broken server. It is dropped only together
with a recognized servers section, where it cannot be a v1 server.

Refs #538

* fix(mcp): read an empty OpenCode 2 servers section

An empty mcp.servers object was not recognized as OpenCode 2's section,
so import listed servers and timeout as broken servers instead of
finding no entries. The dashboard help for disabled also still said only
the switch is written, which is no longer true for opencode.

Refs #538

* fix(mcp): keep an OpenCode endpoint switch next to its own fields

The kept switch was matched by the whole native entry, so a field only
OpenCode reads, such as a per-server timeout, made the next sync drop the
type and command or url once the global server left the config, and
OpenCode 2 then ignored the switch. Match the fields Skillshare manages.

Refs #538

* fix(mcp): show the OpenCode endpoint switch in the dashboard preview

The dashboard renders a project's server alone, so the preview never saw
the same-named global server and showed the bare switch that OpenCode 2
ignores, while sync wrote the switch with the endpoint. Resolve the
global server the way sync does.

Refs #538

* fix(plugin): read OpenCode 2's version output

OpenCode 2 prints "opencode v2.0.27" for --version, while OpenCode 1
prints only "1.19.0". The adapter trimmed a leading "v" only, so plugin
sync to OpenCode 2 failed as an unrecognized version and imported
updates were skipped as if OpenCode 1 had no update command. Read the
last field, and decide the update check with the same parser.

Refs #538
2026-10-12 23:57:40 +08:00
Willie b9563ea4e8 feat(memory): add host-owned review and verification workflows (#539)
* feat(memory): add host-owned review and verification workflows

Share scoped memory tools between Dashboard and Memory, and route retrospective and Dreaming requests through the built-in skill. Keep agent execution in the host, require approval for reviewed edits, and preserve versioned CLI write safeguards.

Confirm scope before exposing prompts and include the resolved root in recent memory responses. Update provenance guidance and localized documentation, with regressions for unavailable scope and protected review prompts.

* fix(memory): distinguish trial copy controls for screen readers

Include the localized step title in each trial copy control accessible name while keeping the short visible label. Update the existing clipboard test selector for the distinct names; no new test or local test rerun was requested for this label-only fix.

* fix(memory): point connection hints to the trial tools

Replace stale verification-prompt directions in all eleven locales with the current Memory tools and Try across sessions labels. Interpolate the existing localized labels so future naming changes remain aligned.

No new tests or local test reruns for this copy-only correction, as requested by the maintainer.
2026-10-12 22:34:05 +08:00
Willie 8213a4a385 feat(ui): make shared memory easier to find and read (#537)
* feat(ui): surface existing shared memory on the dashboard

Make shared notes visible through recent updates and direct links, with file timestamps and an introductory state for untouched starters. Keep the dashboard quiet until memory is configured, and preserve access to existing notes when starter files are missing.

* fix(memory): bound dashboard reads and remember preview expansion

Count vault entries from metadata and read at most the two starters and four recent candidates instead of hashing the full vault on dashboard load. Keep incomplete starter states accessible and refresh the overview after source and instruction changes. Remember preview expansion in browser storage and raise the collapsed height to 360px.
2026-10-12 17:34:33 +08:00
Willie 19cdc1c72e fix(plugin): let an imported Claude plugin update to what its marketplace holds (#536)
An imported plugin's native marketplace decides the version an update installs, so the update check now applies only to bindings with a reviewed source and an import records the version installed. Importing an imported plugin again also records the installed version; a source binding keeps the version Skillshare reviewed.

Refs #535
2026-10-12 17:04:52 +08:00
Willie 99c7911bf3 feat(plugins): update a batch of plugins over one stream (#534)
Updating several plugins from the dashboard was slow: each plugin took a
preview and an apply round trip, apply previewed again, every plugin cloned
its source on its own, and config reloaded after each one.

The dashboard now sends the picks to POST /api/plugins/update/stream. The
server holds the plugin lock once, previews each plugin once and applies that
plan, lets plugins from one source share a single read of it, and reloads
config once at the end. POST /api/plugins/update/stop lets the plugin under
way finish and starts no other; a closed page does not cut an install short.

The check-to-update safety moves to the server: an Agent whose source moved
since the check, or that no check showed, installs nothing. npm packages in
Pi are exempt, since Pi installs whatever npm names latest; the old
client-side comparison flagged them as moved on every update.

The update dialog follows the Skills update page, shows per-Agent old -> new
versions, fails rows a broken stream never finished, and lists every plugin
with its Agent and version when all are current.

Release-As: 0.25.6
2026-10-12 16:09:12 +08:00
Willie 111dd71a53 fix(sync): let prefixed naming keep skills whose name differs from their folder (#531)
Prefixed naming rewrites name: in every copy to the entry name, so the
target folder and name: always agree. Requiring name: to equal the source
folder only skipped skills it could sync: a tracked repo that is itself a
skill lives in _<repo>, so its root SKILL.md could never pass.

Only standard naming now requires the match; name format rules still
apply to both. Sync lists the skills named this way, so a user looking
for _ooux in the target finds ooux.

The dashboard's skipped-skills hint always blamed folders that end in the
same name, even when skills were skipped for their name. The diff API now
reports naming skips apart from collisions and the hint names the cause.

Refs #530
2026-10-12 13:59:59 +08:00
Willie 5338d6a4c1 feat(ui): check plugin updates on load and update them in one batch (#527)
* feat(ui): check plugin updates on load and update them in one batch

Updating plugins took four steps each: press Check updates, press Update
on one row, review a second preview, apply, then check again because the
list did not know the update had happened. Six plugins meant six rounds.

The page now checks once in the background when it opens, so rows show
their new version without a click. The header's Check updates button
becomes "Update N" when something is behind and opens one dialog: every
plugin with an update is ticked, a plugin on several Agents is one row
whose Agents can be left out, and imported installs Skillshare cannot
update sit apart. One button runs them in turn through the existing
preview and apply, so no new API is needed. Progress, results, errors
and per-Agent retry stay in the same dialog, and the applied results
update the page state, so nothing has to be checked again.

The count and the dialog both come from what is still behind after the
last check, limited to Agents that can be updated from here, so a
partial batch is not offered again and Agents with check but no update
(OMP, Antigravity CLI, Grok) do not inflate the count.

* fix(ui): keep batch plugin updates to what the user reviewed

Codex review of #527 found four ways the batch update could act on
something other than what the dialog showed:

- A source that moved between the background check and the click was
  installed straight from the fresh preview, so a newer version or new
  components went in unreviewed. The batch now compares each fresh plan
  with the version and commit the check showed, applies nothing when
  they differ, says so on the row, and runs a quiet check so the next
  dialog shows the new version for review.
- A retry of the failed Agents that failed again lost that subset, so a
  third try re-sent Agents that had already updated. A failed row now
  always keeps the Agents it failed on.
- An update the server skipped (its marketplace gone, say) counted as
  done: the row said updated and the plugin left the update count.
  Skipped outcomes now keep the Agent pending and show as skipped.
- A plugin removed after the check stayed in the update count and
  failed with "not found" when picked. Pending updates now drop any
  plugin or Agent binding the inventory no longer has.

* fix(ui): show a manual check's result and each Agent's own version move

Codex's second review of #527 found two gaps:

- A check the user pressed that found nothing to update from here (only
  imported installs, or everything current) opened nothing, so the
  "update inside the Agent" guidance and the up-to-date result never
  showed. The check now reports whether it came back, and a check the
  user asked for opens the dialog; the background check on load stays
  quiet.
- A plugin on several Agents showed one version move built from the
  first Agent's installed version and the first new version found, which
  can mix Agents that come from different sources. The row now shows one
  move only when every Agent makes the same one, and each Agent's own
  move otherwise.

* fix(plugin): carry the source digest on a check so a batch update can tell it moved

Codex's third review of #527: a local-directory plugin has no commit,
so when its files changed again after the check while the manifest
version stayed the same, the batch's version and commit comparison
passed and the fresh preview was applied without the user seeing it.

The check copied the installed binding, so its digest was the one
already installed, not the source's. It now sets the discovered digest
alongside the version and commit, as the update preview already does.
The dashboard keeps that digest with each update it found and refuses
to apply when the fresh preview's digest differs, marking the row as
changed since the check. An npm package has no digest and is unaffected.

* fix(ui): list blocked check results and drop unticked Agents from the batch

Codex's fourth review of #527:

- A check can mark an Agent blocked when the updated source no longer
  offers a compatible plugin for it. The update dialog listed only
  updatable, native-check and up-to-date results, so it hid the block
  and could say everything was current. Blocked results now get their
  own section with the reason, and the empty state says nothing can be
  updated from here rather than that everything is up to date.
- An Agent unticked after the check stayed in the pending updates;
  picking it produced a no-op that the batch reported as updated.
  Pending updates now skip bindings with sync off, as the row's own
  behind count already does.

* fix(ui): count partly skipped plugins as updated and surface a failed re-check

Codex's fifth review of #527:

- When one Agent of a plugin updated and another came back skipped, the
  row was marked skipped, the summary counted no update, and the Agent
  that did update was not shown. A row now names which Agents updated
  and which were skipped, and the summary counts any plugin with an
  updated Agent.
- After a source moved since the check, the follow-up check ran quietly,
  so if it failed the stale update stayed offered and every retry hit
  the same skip. The moved plugin now leaves the found updates, and the
  follow-up check reports its failure like any check the user started.
2026-10-11 22:19:14 +08:00
Willie 0792d0e1a2 fix(config): keep the comments of a key the command removed (#528)
A comment on a key the command removed was lost. It now sits in front of the next surviving key, or on the last entry of its section, in source order. Adds the config_keeps_comments runbook.

Refs #525
2026-10-11 22:02:55 +08:00
Willie 551a1546df fix(config): keep the comments of config.yaml when a command writes it (#526)
Every writer re-encoded the in-memory struct, so a command such as target --mode copy dropped every comment. marshalConfig now takes the file being replaced and copies its comments onto the keys that are still present; list items are matched by name, and a read error fails the save. registry.yaml goes the same way.

Fixes #525
2026-10-11 21:21:31 +08:00
Willie 89f4b9425a refactor(reconcile): share the lock pin mover, skip adoption when no record is gone (#524)
Cleanup after #523: install.MoveLockPins shared by skillmove and the project reconcile; one server store reload and reconcile dispatcher; adoption skips the source walk when no plain record's directory is missing.
2026-10-11 04:19:32 +08:00
Willie 3a5da6b529 feat(sync): adopt the install record of a skill moved by hand (#523)
After a plain mv, only install or Install missing moved an installed skill's record; sync now adopts it too (CLI global, CLI project, dashboard), never prunes, and skips on --dry-run. A hand-moved project skill keeps its lock pin; a same-name nested skill never takes a live record; metadata load failures are sync warnings.
2026-10-11 03:50:58 +08:00
Willie f1ca009de3 fix(move): recompute collisions after a refusal, report an unreadable .skillignore, preflight batch dry-run --into (#522)
Three follow-ups from the review of #520: a name refused for a
collision no longer makes an unrelated name in the same batch fail
for a path that will not exist, a .skillignore that cannot be read
after the rename is reported instead of silently skipped, and the
batch install --dry-run path runs the same --into preflight as the
real install.

Closes #521
2026-10-11 01:04:53 +08:00
Willie bf1731dec4 feat(move): move installed skills and folders without losing their records (#520)
Add skillshare move and POST /api/resources/batch/move so skills and
whole folders can change folder inside the source while keeping their
.metadata.json entry, accepted audit findings, lockfile pin, config
group, .gitignore and literal .skillignore lines. Moving by hand lost
the record and made the next install start over. install --into now
refuses a folder inside a skill or a tracked checkout.

Refs #510
2026-10-11 00:43:47 +08:00
Willie dcffa2ec5e fix(install): keep a moved skill's record instead of reinstalling it (#513)
* fix(install): keep a moved skill's record instead of reinstalling it

Install records are keyed by path, so a skill moved with mv shows up as
missing. "Install missing" (and bare `skillshare install`) then cloned it
again at the old path, leaving two copies that both sync to targets, and
reconcile could hand the new copy's record to the moved one.

A gone record now follows an unrecorded directory of the same name whose
files match the recorded hashes, when exactly one such directory exists.
Install-from-config skips such entries and reconcile re-keys the record,
so the moved copy keeps its source. Same-name skills with other files are
left alone, so an update can never overwrite them from that source.

Refs: #510

* fix(install): keep audit acceptances and unreadable links out of moves

Re-keying a moved skill went through Remove, which also drops the audit
findings accepted for the old path, so an unchanged skill could block or
prompt again on its next update. MoveEntry now carries them to the new key.

A record below a first-level source link that cannot be read looks gone
to Lstat, so an identical copy elsewhere could take its record while the
link was down. Such records are no longer matched as moved.

* fix(install): drop legacy-path acceptances and respect lock pins on move

A legacy basename key such as `demo` with group `old` keeps its accepted
audit findings under `old/demo`. MoveEntry removed only the key's own
acceptances, so the old path kept them and a later skill placed there
would inherit them without a new --force. It now deletes the old path.

When the lockfile pins another commit than the moved copy holds, treating
the copy as the moved skill re-keyed its stale record and let the project
lock be rewritten with the older commit. Such copies no longer count as a
move, so the pinned commit is installed as before.

* fix(install): search for moved skills only where reconcile looks

The search for a moved copy walked into installed skills and tracked
checkouts, which reconcile skips. A copy found there made install skip the
skill, then reconcile never saw the copy and pruned the record, and in
project mode the config entry with it. The search now stops at the same
boundaries.

* fix(install): never take a recorded skill as another's moved copy

Two identical skills can be recorded at different paths. When one of them
went missing, the other matched its hashes and was taken as its moved
copy, so install skipped the missing skill and reconcile pruned its record.
A directory with a record of its own is now never a move destination.

* fix(install): derive moved-copy paths from the resolved source root

sourcewalk.WalkDir reports paths under the resolved source root, but the
moved-copy search took them relative to the unresolved one. With a linked
source (or Windows short temp paths) every path came out as ../..., so the
installed-skill boundary never matched and a copy inside an installed
parent could be taken as the move, after which reconcile pruned the record.
The search now resolves the root first, as reconcile does.

* fix(install): move a record only to one copy of what the config declares

Three cases still let a move drop or misplace a record:

- In project mode the manifest can point a skill at a new source while a
  copy of the old one exists elsewhere; that copy was taken as the skill,
  so the new source was never installed. The record's source and branch
  must now match the config entry.
- A git checkout reconcile treats as tracked could be accepted before the
  tracked boundary was checked; the boundaries now run first.
- With two identical copies, reconcile re-keyed the record to whichever
  it walked first. Both reconcile and the install-from-config search now
  require exactly one destination, and reconcile decides after the walk.

* fix(install): clear accepted findings left at a move destination

A removed legacy grouped entry can leave findings accepted for its path
behind. MoveEntry only wrote the destination when the moved skill had its
own acceptances, so a skill moved into that path inherited findings it
never accepted. The destination is now cleared before the transfer.

* fix(install): follow a moved skill only while its declaration still matches

The install-from-config search compared source and branch but not the
tracking mode, so a skill switched to tracked was satisfied by an old
plain copy and never cloned.

Project reconciliation, which other installs also run, inferred moves
without looking at the manifest at all. A record whose declaration now
names another source could follow a copy of the old source, after which
the manifest entry was rewritten to that old source. Moves in project
mode now require the declaration to still match the record, through the
same MatchesDeclaration check the install path uses.

* fix(install): infer no move while a source link is unreadable

An unreadable followed source link can hide a second identical copy, so
a copy that looked unique could take the record arbitrarily. Reconcile
and the install-from-config search now infer a move only from a complete
inventory; reconcile already keeps the gone record in that case.

* fix(install): infer no move when the source walk hit an error

Both walks dropped traversal errors, so an unreadable directory that hid
a second identical copy left the visible one looking unique and it could
take the record. Any walk error now counts as an incomplete inventory,
like an unreadable source link, and no move is inferred from it.

* fix(install): keep a record whose move waits on a failed walk

When a walk error deferred a move, the walk still counted as complete, so
pruning dropped the gone record along with its accepted audit findings
and it could never follow the copy. A failed walk that deferred a move
now marks the inventory incomplete, as an unreadable source link does;
a failed walk with no move candidate keeps the old behavior.

Also correct InstallFromConfig's comments, which still said reconcile
runs only after an install.

* fix(install): move a moved skill's gitignore rule in project mode

Project reconcile added the managed ignore rule for a moved skill's new
path but left the old one, and the gitignore update only ever adds. If
the old path later held source-controlled files, Git kept ignoring them
without a word. Reconcile now reports the paths records moved away from,
and project reconcile removes their managed rules before adding the new.

* test(config): skip the unreadable-directory reconcile test on Windows

Windows does not stop reading a directory created with mode 0, and
os.Geteuid reports -1 there, so the root-only skip let the test run and
fail on the Windows runner.

* fix(install): count a candidate that cannot be hashed as an incomplete search

MovedEntryKey dropped a same-name candidate whose files could not be
hashed, so a readable identical copy beside it looked unique and could
take the record. It now returns the hash error, and reconcile and the
install-from-config search treat it like a walk error: no move is
inferred from that scan, and reconcile keeps the record waiting.
2026-10-10 23:14:52 +08:00
Willie 81974cf350 fix(resource): never skip a hidden-named agents source root (#515)
AgentKind.Discover and the skill detail file listing skipped every
hidden directory, including the walk root itself. The `!= "."` guard in
agent.go never fires because walkRoot is an absolute resolved path, so a
sources.agents directory such as ~/.agents discovered zero agents.

Exempt the walk root (`path != walkRoot`), matching the other walks in
the repo. Nested hidden directories are still skipped.

Related to #511 (same class as #514).
2026-10-10 22:03:53 +08:00
Willie d2f5936be9 fix(sourcewalk): never skip the source root as a target dotdir (#514)
SkipTargetDotDir only exempted dotdirs directly under the walk root, never
the root itself. ProjectTargetDotDirs() always contains ".skillshare", so a
source whose basename is a target dotdir (".skillshare", ".claude", ...) had
its first Walk callback return SkipDir. Discovery then saw an empty source:
`skillshare list` returned [] and sync pruned previously managed links.

Exempt path == walkRoot in the shared helper so both callers
(sync discovery and resource skill discovery) are fixed. Nested target
dotdirs are still skipped and first-level ones stay visible.

Fixes #511
2026-10-10 21:43:35 +08:00
Willie 2ce46834e2 fix(collect,install): collect only real skills; let --track take a local git path (#508)
* fix(collect,install): collect only real skills; let --track take a local git path

collect treated every directory in a target as a local skill, so a scratch
folder was listed and `collect --force` copied it into the source. A folder
now counts only when it holds SKILL.md; the dashboard scan shares
FindLocalSkills and the dashboard collect request rejects such a folder.

`install /path/to/repo --track` failed with "requires a git repository
source" while file:///path worked. A local path that is a git repository
root is now normalized to the same file:// clone URL for --track (also past
the --branch guard); other local folders get an error that names file://.
Help text, docs (English and i18n) and the built-in skill reference follow.

* fix(install): keep the drive letter in file:// clone URLs for local --track paths

A Windows path such as C:/repo produced file://C:/repo, which makes C: the
URI authority instead of part of the path, so a tracked install from a local
Windows repository could clone the wrong location. Prefix the missing slash
so the URL is file:///C:/repo; POSIX paths are unchanged. Also cover the
dashboard install endpoint with a local-path --track test.

* fix(collect,install): keep UNC file URLs intact; require SKILL.md to be a file

The drive-letter fix lived in the helper shared with explicit file:// URLs,
which rewrote file://server/share/repo into a local path. Keep the slash
logic in the local-path --track case only and leave explicit URLs as parsed.

A subdirectory named SKILL.md passed the existence check, so a scratch folder
holding one was still collected. Both collect entry points (scan and the
dashboard request) now share sync.HasSkillFile, which rejects directories.

* fix(install): escape the local path in the file:// URL used by --track

Git decodes a file:// URL, so a repository path containing %20, %2F, a space
or # was cloned from a different path than the one IsGitRepo accepted. Build
the URL with net/url so those characters are escaped; drive-letter and POSIX
paths keep their form. The integration test now installs from a directory
whose name contains a space and a literal %20.

* fix(install): accept a local bare repository path for --track

A bare repository has HEAD and objects at its root but no .git entry, so the
local-path check rejected it while file:///same/path cloned fine. Add
IsLocalGitRepo (working tree or bare) and use it for both the --track
normalization and the --branch guard.

* test(install): build the bare repo path from a Windows file URL correctly

makeRemote returns file:///C:/dir on Windows; trimming only file:// left
/C:/dir, which filepath.Abs turned into D:\C:\dir, so the bare-repo test
failed on the Windows runner. Drop the slash before the drive letter.

* fix(install): refuse a --track local source that is the install destination

--force removes the tracked destination before cloning. When the local path
given to --track is that same directory, the removal deleted the only clone
source (and any unpushed work) before the clone failed. Compare the two with
os.SameFile, which also covers symlinked paths, and stop before touching the
destination.

* fix(install,collect): keep a tracked source inside its destination; reject special SKILL.md files

--track --force removed the destination before cloning, so a local source
nested below it (or given as an explicit file:// URL) was deleted along with
any unpushed work, and the clone then failed. Resolve the source and refuse
when it is the destination or lies under it, for dry runs too.

HasSkillFile used a not-a-directory test, so a FIFO named SKILL.md counted as
a skill and the collect copy blocked on it. Require a regular file; a
symlinked SKILL.md still resolves through os.Stat.

* fix(install): treat file://localhost URLs as local clone sources

Git clones file://localhost/path from the local filesystem, but the
containment guard only recognised an empty authority, so --track --force
could still delete a source given that way. Accept localhost (any case) as
local.

* test(install): build the localhost file URL from a slash path

On Windows the temp path starts with a drive letter, so prefixing it onto
file://localhost produced file://localhostC:\... instead of a valid URL and
the case failed on the Windows runner. Convert to slashes and add the
separator explicitly.

* fix(install): clone --track --force into a staging dir and swap it in last

--force removed the existing tracked repo before cloning, so any source that
lived under it (plain path, file:///, file://localhost, UNC, symlink or 8.3
alias) was destroyed before git could read it, and a failed clone also lost
the old repo. Clone beside the destination, run the checks and the audit on
the staging copy, then move the old repo aside, rename the new one in and
drop the backup; on any failure the staging copy is removed and the old repo
stays. The pre-removal containment refusal now covers plain local paths only
(the url.Parse special-casing is gone).

* fix(install): refuse a file:// --track source inside the forced destination

The staged swap in 7375400 still lost data for a file:///, file://localhost
or UNC source nested under the destination: the clone succeeded, the old repo
(with the source inside it) was renamed to the backup, and removing the backup
deleted the source's uncommitted work, which a clone cannot carry.

Restore the file URL to local path mapping (now including UNC hosts) so the
containment refusal covers every local spelling, not only plain paths. The
test now keeps an uncommitted file in the source and requires it to survive.

* fix(install): apply the source-inside-destination refusal only to --force

The refusal ran before the --update branch, so `install <tracked repo path>
--track --update` was rejected even though an update pulls the existing
checkout in place and removes nothing. Only the forced replacement can
destroy a source that lives under the destination, so check there.

* fix(install): report a forced clone's root skill under its tracked name

With --force the repo is discovered in the .skillshare-clone-<stamp> staging
directory, and discoverSkills names a root SKILL.md after the directory, so
the CLI summary, dashboard response and oplog listed a random staging name
instead of _foo. Use the tracked name for the root entry.

* fix(install): stage fresh tracked clones so cleanup never hits another installer

A fresh tracked install cloned straight into the destination, and the
deferred cleanup removed it on any failure. If another process created the
same tracked repo between the existence check and the clone, the failing
install deleted that process's checkout. Clone every install into its own
staging directory and rename it in last; a claimed destination makes the
rename fail and only the staging copy is removed.
2026-10-10 01:06:23 +08:00
Willie c638b46096 fix(sync): replace copy-mode copies with links when switching back to merge (#506)
* fix(sync): replace copy-mode copies with links when switching back to merge

Switching a target from copy back to merge left every copy whose entry name
did not change as a real folder. Merge sync treated it as the user's own
folder, so status and doctor kept saying "run skillshare sync" while sync
changed nothing, and only `sync --force` turned it into a link.

Copy mode already replaces merge-mode links with copies on its own, so merge
sync now replaces a folder the manifest records with a checksum (a copy-mode
copy) with a link, without --force. Folders the manifest does not record, or
records as a link, stay untouched. The pre-sync target backup is unchanged.

The CLI and dashboard diffs name these entries as copies sync replaces
instead of suggesting --force, and the sync-modes docs describe the switch.

* fix(sync): keep copy-mode copies the user edited when switching to merge

The first version replaced any folder the manifest recorded as a copy-mode
copy. The manifest checksum is only a record, so a copy edited since copy
mode made it was deleted too, and `sync --json` and project-mode dashboard
sync take no pre-sync backup to restore it from.

Replace the folder only while it still hashes to the recorded checksum
(Manifest.OwnsCopy). An edited copy is kept and `sync --force` still
replaces it; diff keeps suggesting --force for it. A copy whose name: was
rewritten by prefixed naming, or that was made with custom ignore patterns,
no longer hashes to the source checksum and is kept the same way.

* fix(ui): say a copy-mode copy in a merge target is replaced by a link

The Sync page picks a row's text from the diff item's action alone. The
copy-mode copy that merge sync now replaces comes from the server as an
update, which the page worded as "Link is broken or points elsewhere. Will
be relinked" even though no link exists yet.

Match the server's reason and show "Copy made by copy mode. Will be replaced
by a link" instead, in all eleven dashboard languages.

* docs(sync): say only unedited copy-mode copies are replaced on merge sync

The sync-modes paragraph promised that copies copy mode made are replaced
with links without --force. Merge sync keeps a copy that no longer hashes to
the checksum copy mode recorded, so an edited copy, or one whose name: copy
mode rewrote for prefixed naming, still needs --force.

State the three cases in sync-modes and qualify the diff Modified and Local
override labels, in English and the ja, ko, zh-Hans and zh-Hant copies.

* fix(sync): keep the copy until its replacement link exists

Replacing a copy-mode copy removed the folder first and then created the
link. If creating the link failed (no symlink or junction rights on a
Windows or network filesystem), sync returned an error with the skill gone
from the target. This path now also runs on a plain sync, not only on
--force.

Park the folder beside the entry, create the link, then delete the parked
folder; on failure the folder is moved back. A parked folder left by a run
that died mid-way may be the only copy, so the next sync stops with a message
instead of deleting it. The manifest also forgets the copy-mode mtime and
naming records of an entry that is now a link.

* test(sourcefs): register replaceWithLink's target writes in the raw-write ratchet

The previous commit moved the merge-mode folder replacement into
replaceWithLink, which the ratchet reports as new raw os calls and the old
os.RemoveAll allowance as stale. All of them act on the sync target, not the
skills source, like the neighbouring entries.

* fix(sync): bound the parked copy's name and document the naming exception

The folder parked while a link is created was named after the entry, so an
entry near the 255-byte filename limit (a long flattened nested skill)
could not be moved aside and its replacement failed. The parked name is now
a fixed-length hash of the entry name.

sync-modes also says what happens when `target_naming` changes in the same
sync: the copy gets a new entry name and the prune pass removes it as an
orphan whether or not it was edited, as it did before this change.
2026-10-10 00:13:29 +08:00
Willie b0d87fba28 feat(sync): make prefixed target naming misconfigurations visible (#503)
* feat(sync): make prefixed target naming misconfigurations visible

Prefixed naming (#497) works, but several of its failure modes only showed up
at sync time or gave advice that cannot be followed:

- A collision that only prefixing creates told the user to "rename one in
  SKILL.md", which a tracked skill cannot do, and never said the prefix caused
  it. The hint now names the prefixing and points at renaming the other skill or
  re-tracking the repo with --name.
- `target add` silently wrote mode: copy when the inherited naming is prefixed.
  The CLI now says so.
- A target that resolves to prefixed outside copy mode was reported only by
  sync. status, doctor and target list (plain and JSON) now flag it, and the
  sync/validation and `--skills=true` errors say how to fix it.
- The dim "duplicate skill names (isolated by ...)" line printed on every sync
  even when nothing was wrong, and before a failing sync. Isolated duplicates
  are not a problem, so it is dropped; the website docs still showed an
  outdated form of this output and are corrected in all locales.

* fix(status): flag prefixed-outside-copy in the TUI list and status --json

Codex review of #503 found two surfaces the new check missed:

- The interactive `target list` (the default on a terminal) never read the
  item's naming error, so a prefixed + merge target looked healthy there. The
  list row now marks the mode with "!" and the detail panel shows the fix.
- `status --json` (global and project) builds its target entries directly, so
  scripts got no signal. Each target now carries a `warning` string, matching
  `target list --json`.

* docs(troubleshooting): sync no longer prints a message for isolated duplicates

The collision troubleshooting tip still said sync shows an info message when
filters isolate duplicate skill names. That line was dropped, so the tip now
says sync prints nothing about them. Codex review of #503 flagged the stale
English page and its four translations.

* fix(config): point managed-project naming errors at the project settings

The "set mode: copy on the target" hint is wrong for a target a global
`projects:` entry expands into: it cannot be edited with `target` or under
`targets:`, and the mode lives in the project's skills settings. The hint now
depends on whether the target belongs to a managed project (Codex review of
#503).

The naming/mode check becomes a TargetConfig method so status, doctor and
target list share it, which also lets the hint see the project root. Along the
way: doctor reuses it, the TUI list renders the mode once, and
reportTargetAdded takes a flag instead of comparing a mode string.

* fix(config): hint the project settings for managed-project naming errors

ProjectNamingError (the config-load check for a global `projects:` entry) still
returned the bare naming error, so a project that inherits prefixed with a
non-copy mode got no fix. It now names the project's skills settings, like the
expanded targets do. The built-in skill's sync reference also mentions that
status, doctor and target list flag the misconfiguration before sync.

Found by the second /code-review pass on #503.

* fix(sync): detect prefix collisions from the colliding paths

reportCollisions decided a collision came from prefixing when its name was not
also a duplicate in the source. A tracked skill that joined two ordinary
skills already sharing the name was therefore sent to the "rename one in
SKILL.md" hint, which a tracked skill cannot follow (Codex review of #503).

A collision on a prefixed target that includes a tracked path now gets the
tracked-skill hint, whatever else shares the name. The wording no longer claims
prefixing renamed the skill, since a name that already starts with the repo
name is unchanged.

* fix(sync): use discovered repo metadata for the tracked-skill hint

The collision hint treated any path segment starting with "_" as a tracked
repo, so an ordinary skill under a folder such as org/_drafts (no .git) was
told it could not be renamed and to re-track a repo that does not exist. It now
uses DiscoveredSkill.IsInRepo, the same signal prefixed naming uses.

The prefixed docs also gain the fix for a target a global `projects:` entry
expands into (projects.<root>.skills.mode: copy), since `target` cannot edit
those. Both from Codex review of #503.

* fix(sync): keep both collision hints when targets disagree

The same name can clash on a prefixed target with a tracked skill and on
another target between ordinary skills only. The prefixed flag was OR-ed across
targets, so the second case lost its rename/filter hint. Track both kinds
separately while grouping, and print each hint that applies (Codex review of
#503).

* fix(sync): keep the rename hint inside a mixed collision group

When a tracked skill and two ordinary skills share a prefixed name, re-tracking
the repo fixes only the tracked one; the ordinary pair still clashes and is
excluded. The group now also shows the rename/filter hint when at least two of
its participants are not tracked (Codex review of #503).

* fix(sync): offer filters in the tracked-skill collision hint

Two tracked skills of the same repo with the same name collide after
prefixing, and neither renaming a tracked skill nor re-tracking the repo
separates them. The hint now also names include/exclude filters, so every
collision on a prefixed target has an actionable remedy (Codex review of #503).

* docs(sync-modes): limit the target list claim to targets under targets

target list builds from the config without project expansion, so it cannot
flag a target a `projects:` entry expands into; status and doctor can. The
prefixed paragraph and the built-in skill reference said all three flag it
(Codex review of #503).
2026-10-10 00:00:23 +08:00
Willie ec2a8ab3aa fix(sync): close gaps found while simulating the prefixed naming switch (#504)
* fix(sync): close gaps found while simulating the prefixed naming switch

Simulating the #497 reporter's switch from merge + flat to copy + prefixed
turned up four problems:

- `target X --mode copy --target-naming prefixed` applied only --mode, since
  the setter dispatch returned after the first flag. Both flags given
  together now check the final pair, then apply in an order that keeps the
  saved pair valid (copy mode first, any other mode last).
- Skipped skills were only counted, so a prefixed name over 64 characters
  vanished with "1 skill(s) skipped". Sync now names the first five with
  their reason; the cap keeps huge sources from flooding the output.
- In copy mode a merge-mode link to the skill was not a managed entry, so
  when a local folder held the new name the old link was pruned and the
  skill left the target, contrary to the documented "old managed entry is
  preserved". Copy mode now migrates such links like managed copies.
- PATCH /api/targets validated the mode/naming pair against the current
  skills state, so disabling skills while changing mode was rejected. It now
  checks the state the request leaves the target in.

* docs(target): drop stray space in the Japanese mode+naming sentence

* fix(diff): save paired target settings once and report kept legacy entries

- `target X --mode M --target-naming N` ran the two setters, each saving
  the config, so a failed second save left half the change behind (Codex
  on #504). Both fields are now set and saved once, after the final pair
  is checked.
- When a local folder holds the name a skill's naming now gives it, sync
  keeps the skill under its old entry and leaves the folder alone, even
  with --force. diff still called the folder a "local copy" and suggested
  `sync --force`, which does nothing for it. CLI and Web API diff, copy
  and merge mode, now report the folder and the entry the skill stays at.

* fix(sync): migrate only links the manifest records as merge-made

Copy mode treated any link to the skill's source at an old naming's name as
the entry merge mode left behind, so a user's own link was renamed and
replaced with a copy (Codex on #504). A link now migrates only when the
manifest records it as "symlink", which merge mode writes for links it makes;
other links stay as the user left them. The value is now a named constant.

* fix(diff): mark kept legacy entries as unaffected by force

Two problems Codex found on #504:

- The Web API diff sent a local folder on a skill's new name as a "skip",
  which the dashboard shows as an edited entry that Force replaces. Sync
  keeps that folder and the legacy entry even with force, so the preview
  promised a change that never happens. It is now its own "kept" action,
  shown with its reason and not counted.
- `target X --mode M --target-naming N` printed the target's own naming as
  the old value, so a target inheriting standard or prefixed showed
  "flat -> ...". It now prints the naming the target had in effect.

* fix(diff): keep kept-name conflicts out of the sync count

Two problems Codex found on #504:

- A local folder on a skill's new name counted toward "N to sync" and
  fell through to the plain `skillshare sync` hint, though sync changes
  nothing there until the folder is renamed or removed. Such rows now
  carry a kept flag, count as neither sync nor local, and print their own
  hint; a target whose only row is kept still shows that row.
- The dashboard grouped targets by row name and icon only, so two targets
  keeping the same skill at different old entries merged and showed one
  target's old location for both. The row detail is now part of the key.

* fix(diff): do not list a kept local folder's files as deletions

A kept row (a local folder on a skill's new name) still pointed at that
folder, so `diff --stat` and `--patch` listed each of its files as removed,
though sync never touches the folder (Codex on #504). Kept rows no longer
point at a target directory, so no file-level diff is built for them in the
CLI or the TUI.

* fix(diff): report kept skills consistently in every diff view

An exhaustive review of the kept-row change (prompted by two Codex findings
on #504) found the same gap in several places:

- A target whose only rows were kept was summarized as "in sync", and mixed
  summaries left it out of every count. Such targets now count as "N kept
  under old name" with a warning mark, and the scan progress counts them.
- Plain output and the TUI hid where the skill stays; rows now read
  "name (stays at old-name)". The TUI also lacked the kept Next hint and
  called kept-only targets "differs"; it now matches the plain output.
- `diff --json` emitted kept rows as "remove", indistinguishable from
  local-only folders; it now emits "kept", as the Web API does. The diff
  docs' JSON example also showed actions the CLI never emits.
- Renamed rows whose old entry is a merge-made link showed a bogus file
  stat, because the link was walked as a file; a link now stands for the
  folder it points to.
- A dry-run migration printed "Would copy" to the old path; it now names
  the new one.

* fix(sync): stop reporting kept folders and skipped skills as pending work (#509)

Three reports told users to do something sync cannot do, or hid what it
did:

- Copy and merge sync kept a folder the user made on a skill's name but
  counted it as "up to date" with no hint. Sync now counts it as local and
  adds one capped line per target naming the folders and saying that
  `sync --force` replaces them. The line is added in SyncSkillTarget, so
  the dashboard gets it too.
- status and doctor counted skills that sync skips on purpose (invalid
  names under standard or prefixed naming, name collisions) as "not
  synced — run skillshare sync". Running sync can never add them, so the
  advice looped. They now compare against the skills sync places.
- diff marked a target whose only items were local-only folders as "to
  sync" and `synced: false`. Those items are still listed, but they no
  longer make the target pending.

* fix(target): apply every setting flag, require flag values, report no-ops as unchanged (#507)

* fix(target): apply every setting flag, require flag values, report no-ops as unchanged

`skillshare target <name>` dispatched --mode, --agent-mode and
--target-naming through an if-chain that returned after the first match,
so `--agent-mode copy --target-naming prefixed` applied one setting and
silently dropped the other. Filters, --skills and --dry-run had the same
early-return problem.

- Check --mode, --agent-mode and --target-naming as a whole and save
  once, in global and project mode (one targetSettingsUpdate replaces six
  per-flag functions).
- Reject --mode/--agent-mode/--target-naming combined with filters or
  --skills, and --dry-run without --skills, instead of dropping them.
- A value flag followed by another flag (or an empty string) now reports
  "<flag> requires a value"; `--target-naming -p` used to report an
  invalid naming '-p'.
- A value the target already has is reported as unchanged and the config
  is not rewritten.

* fix(target): leave the skills pair alone on agents-only changes

A target with skills off may keep target_naming prefixed with a merge
mode it cannot sync; validateGlobalTarget and validateProjectTarget only
check that pair for enabled skills. The combined settings update ran the
pair check for every setting, so `target claude --agent-mode copy` was
refused on such a target although the old agent-mode handler never looked
at skills settings. Check the pair only when --mode or --target-naming is
given.

* docs(target): show combined settings in --help and the built-in skill

The combined --mode/--agent-mode/--target-naming form is now the way to
move a target to copy + prefixed in one step, but neither the help
examples nor the built-in skill's target reference showed it.
2026-10-09 23:59:47 +08:00
Willie 9c994faa17 fix(frontmatter): read and rewrite SKILL.md that starts with a UTF-8 BOM (#505)
* fix(frontmatter): read and rewrite SKILL.md that starts with a UTF-8 BOM

Windows editors such as Notepad write a BOM before the opening ---. The
lenient readers (ParseSkillName, ParseFrontmatterField/Fields/List, ReadSkillBody)
did not skip it, so under target_naming standard/prefixed the skill had no
readable name and was silently skipped. Strip the BOM in the shared frontmatter
locator and the line scanner, and in the hub, search, audit and sync/context
readers that split lines themselves (utils.TrimBOM).

The writers kept working on the unstripped text, so SetFrontmatterValue
prepended a second block and ToggleFrontmatterFlag/RewriteFrontmatterList
could leave a BOM in the middle of the file. They now split the BOM off and
put it back in front, rewriting only the name value (or flag/list) as before.

* fix(frontmatter): cover the dashboard parser and stop auditing a leading BOM

A consumer sweep after the first BOM commit found two more places that
mishandled a SKILL.md written by Notepad:

- audit reported the leading BOM as a HIGH hidden-unicode finding (U+FEFF is
  in the zero-width rule), so every such skill showed a risk score. A BOM at
  the very start of a file is its encoding mark and hides nothing; one past
  the start is still reported.
- the dashboard's parseSkillMarkdown saw no frontmatter, so the skill editor
  would treat the whole file as body and write a second block on save.

The audit-engine page and its four translations note the exemption.

* fix(ui): keep the BOM when the skill editor saves a SKILL.md

parseSkillMarkdown dropped the BOM so the editor could see the frontmatter,
but SkillEditor writes the composed text back through PUT .../content, so every
dashboard edit silently removed the marker from a Notepad-written file. The
parse result now carries the BOM and composeSkillMarkdown puts it back in front,
matching the Go rewrite paths.
2026-10-09 23:58:56 +08:00