Commit Graph
434 Commits
Author SHA1 Message Date
Willie 0117161c8b Merge remote-tracking branch 'origin/main' into runkids/436-managed-instruction-block 2026-10-07 01:13:44 +08:00
Willie fe4fcba122 Merge branch 'main' into runkids/arch-442-uninstall 2026-10-07 01:07:38 +08:00
Willie 4c6d152abf refactor(uninstall): run the CLI and dashboard through one operation
The uninstall sequence (move-out check, dirty check, move to trash,
.gitignore cleanup, metadata prune) lived in the CLI and in three
dashboard handlers, and the copies had drifted: three different prune
rule sets, no dirty check on DELETE /api/repos, and no metadata prune
at all on DELETE /api/resources, which left the removed skill's entry
and target overrides in .metadata.json.

internal/uninstall now owns the sequence for resolved items. Callers
keep name resolution, prompts, output, HTTP status and oplog.
MetadataStore.RemoveByNames is the single prune rule set: an entry goes
when it is a removed path or lies below one, by key or by Group, plus
the tracked legacy group of a top-level repo.

Behaviour changes, all intended:
- DELETE /api/repos/{name} refuses a dirty tracked repo, or one whose
  git status cannot be read, with 409 unless ?force=true.
- DELETE /api/repos/{name} refuses a followed link whose folder is
  itself a skill, like every other uninstall path.
- DELETE /api/resources/{name} prunes the skill's metadata.
- Removing top-level "foo" no longer prunes "frontend/foo" (CLI), and
  removing repo "_team" no longer prunes the untracked group "team"
  (dashboard batch); both dropped metadata of skills still installed.

Refs #442
2026-10-07 00:13:47 +08:00
Willie 2edf7649fd refactor(update): update tracked repos through one shared operation
Updating a tracked repo is a fail-closed sequence: check for uncommitted
changes, pull, audit what was pulled, reset to the pre-pull commit when the
audit blocks. The CLI had two copies of the pull flow, the dashboard a third,
and the audit gate existed four times (CLI, dashboard, and two in install).
The copies had drifted: the dashboard ignored a failed git status and never
refreshed root-skill metadata, and each copy worded a failed rollback on its
own.

update.TrackedRepo now owns the sequence and returns a structured result;
install.AuditGate is the only gate. internal/update is a new package because
internal/git already imports internal/install. The CLI keeps its spinner,
rows and the "Apply anyway?" prompt, passed in as one Confirm callback; the
dashboard turns the result into its JSON item.

Behaviour changes, all towards the stricter copy:
- the dashboard reports a failed git status as an error instead of pulling;
- the dashboard refreshes root-skill metadata after an update, like the CLI;
- a prompt that fails rolls the update back instead of leaving it applied;
- the rollback-failed and rollback-unavailable texts come from one place.

Accepted findings stay as they were (CLI honours and records them, the
dashboard does neither), now as the explicit AcceptedFindings input. The
"blocked and rolled back" texts are unchanged per caller because the
dashboard matches on them.

Refs #443
2026-10-07 00:10:54 +08:00
Willie ea83840bae test(update): pin tracked-repo update behaviour of the dashboard and CLI
The dashboard's pull flow for tracked repos had no tests of its own, and the
CLI had none for uncommitted changes. Both flows are about to be replaced by
one shared operation, so their current results are pinned first: clean
update, blocked and rolled back, skip-audit, force not remembering findings,
dirty repos, and the pull error hint.

Refs #443
2026-10-07 00:10:54 +08:00
Willie 05ff8c675a feat(extras): prepend and append modes for single-file extras
Tools without an @import directive (codex, gemini, cursor) could only
take a shared instruction file as a symlink or copy, which replaces the
target's own file. Teams that want one shared AGENTS.md plus the user's
own notes in the same file had no way to express that.

prepend and append write the source into the target file as a managed
block fenced by skillshare:extra markers that carry the source path and
a content hash. Sync inserts the block once and then rewrites it in
place, status reports modified when the block body no longer matches
its hash, Collect copies the edited body back to the source, Reapply
rewrites the block from the source, and restore removes only the
block. Lines outside the block, and markers inside code fences, are
never touched. Several managed blocks (and an import line) may share
one target file; the connection validator only rejects a file when a
replacing mode (symlink, copy, merge) is involved.

Refs #436
2026-10-07 00:04:00 +08:00
Willie 2647e79ef9 feat(mcp): require HTTPS for mcp serve off loopback
A non-loopback --http listener accepted a bearer token over plain HTTP,
so anyone on the path could read the token and the skills. Off loopback
it now also needs --tls-cert and --tls-key and serves HTTPS; loopback
stays plain, so a TLS-terminating proxy in front still works.
2026-10-06 16:08:44 +08:00
Willie 64898e145b fix(mcp): keep migration notices off stdout and require closed frontmatter
The first command after an upgrade reports legacy data it moved. For
mcp serve that notice landed on stdout ahead of the handshake, so it now
goes to stderr like serve's other messages.

A SKILL.md that opens frontmatter but never closes it could still parse,
so serve published a file hosts read as having no frontmatter; it is now
skipped with a warning.
2026-10-06 14:57:49 +08:00
Willie fd4f499be7 fix(mcp): keep serve digests and stdout trustworthy
Size and mtime cannot prove a file is unchanged (cp -p, rsync -t), and a
stale digest makes hosts that verify digests reject the file for good,
so every rebuild hashes every file again; 1,000 skills rebuild in about
140 ms, at most once per refresh interval.

Serve errors now go to stderr: stdout is the protocol stream, and a
startup error printed there is not JSON-RPC.

SKILL.md must begin with its frontmatter, as the Agent Skills format
requires; a later --- block no longer passes.
2026-10-06 14:39:21 +08:00
Willie cc7d802411 feat(mcp): serve skills over MCP with the SEP-2640 Skills extension
Agents on a disposable VM, in a hosted assistant or behind an MCP gateway can
reach an MCP endpoint but not the folder skills sync to, so they got tools and
none of the user's skills (#428).

`skillshare mcp serve` is a read-only server for the skills skillshare already
manages. It declares io.modelcontextprotocol/skills and implements skills/list,
skills/get and resources/read: one entry per skill, its URI from the source
path, the full frontmatter, and every file with a sha256 digest and size.
--target applies that target's selection, and the source is reread at most
every 5 seconds so install, update, enable and disable show up without a
restart; unchanged files keep their digest. stdio is the default; --http
serves Streamable HTTP and needs SKILLSHARE_MCP_TOKEN off loopback.

Skills the extension cannot carry are skipped with a warning instead of being
published incomplete: a name that breaks the Agent Skills rules or differs from
its directory, more than 512 files or 16 MiB, or a nested skill that is not
served. --check lists them without serving. The naming rules moved to
internal/skillpkg so sync and the server share them.

Few clients support the extension yet, so the server also offers list_skills
and read_skill tools. A client that declares the extension does not see them,
so it never gets each skill twice.

sync.SelectTargetSkills is the one place that turns discovered skills into a
target's selection (enabled, include/exclude, frontmatter targets).

Refs: #428
2026-10-06 14:18:16 +08:00
Willie 7d5776ef79 fix(link): show help before loading the scope and accept -g/-p as --name values
link -h loaded the config first, so it failed with "config not found"
without one and, with --project, could create .skillshare files just to
print help. Detect help among the option arguments before loading the
scope. Pass --name as a value flag to parseModeArgs so "--name -g" no
longer selects global mode (#422).
2026-10-06 00:02:28 +08:00
Willie 2d5ab6771c Merge remote-tracking branch 'origin/main' into runkids/419-link-unlink 2026-10-06 00:01:28 +08:00
Willie 1d66193749 Merge pull request #425 from runkids/runkids/422-mode-option-values
fix(cli): preserve mode flags used as option values
2026-10-06 00:00:33 +08:00
Willie 1f4a466eed fix(uninstall): refuse a linked root skill before the dirty preflight
A followed checkout whose root is a skill is classified as a tracked
repository, so a dirty checkout hit the uncommitted-changes preflight and
was told to retry with --force before the linked-root guard ran. Run the
guard right after target validation for every mode so the answer is always
"use unlink".
2026-10-05 23:46:02 +08:00
Willie e329f5d906 fix(cli): preserve mode flags used as option values
Pass value-taking options to the shared mode parser so command values cannot select a scope or create false mode conflicts. Preserve the end-of-options delimiter and reuse it for MCP command arguments.

Cover existing CLI entry points, including audit rules, enable/disable, and backup deletion, with regression tests and document the parser contract.
2026-10-05 23:12:35 +08:00
Willie ad5d5cd09f fix(link): roll back the link when enabling follow_source_links fails
link --enable created the link, logged success, and only then wrote the
config; a failed config write left a link discovery ignores behind a
logged success, and a retry failed with "already exists". Enable before
logging and remove the fresh link when enabling fails so the command,
the filesystem, and the operation log agree.
2026-10-05 23:01:50 +08:00
Willie ecb410ff95 test(link): compare canonical paths for the option terminator case
macOS temp directories live under /var, a symlink to /private/var, so
EvalSymlinks on the created link returns the /private form while the test
compared it with the raw t.TempDir() path. Canonicalize the expected path
as well so the integration test passes on macOS runners.
2026-10-05 22:58:13 +08:00
Willie 9db36b472a Merge remote-tracking branch 'origin/main' into runkids/419-link-unlink
# Conflicts:
#	internal/server/handler_skills.go
#	skills/skillshare/references/TROUBLESHOOTING.md
#	website/docs/reference/commands/ui.md
#	website/docs/reference/targets/configuration.md
#	website/i18n/ja/docusaurus-plugin-content-docs/current/reference/commands/ui.md
#	website/i18n/ja/docusaurus-plugin-content-docs/current/reference/targets/configuration.md
#	website/i18n/ko/docusaurus-plugin-content-docs/current/reference/commands/ui.md
#	website/i18n/ko/docusaurus-plugin-content-docs/current/reference/targets/configuration.md
#	website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/reference/commands/ui.md
#	website/i18n/zh-Hans/docusaurus-plugin-content-docs/current/reference/targets/configuration.md
#	website/i18n/zh-Hant/docusaurus-plugin-content-docs/current/reference/commands/ui.md
#	website/i18n/zh-Hant/docusaurus-plugin-content-docs/current/reference/targets/configuration.md
#	website/src/data/featureMap.ts
2026-10-05 22:49:25 +08:00
Willie e7eecffeab Merge remote-tracking branch 'origin/main' into runkids/274-follow-source-links 2026-10-05 21:32:38 +08:00
Willie 97446744ff feat(omp): add native Oh My Pi support (#423)
* feat(omp): add native Oh My Pi support

Support OMP skills, instructions, MCP, code hooks and guarded extension selection using its verified native contracts rather than Pi aliases.

Add reviewed marketplace plugin installation, cache-preserving scoped removal and fresh-cache reinstall. Align Pi/OMP extension cards and localize inventory guidance while keeping native ownership and read-only constraints.

Include regression coverage, isolated E2E runbooks, documentation and verification policies. Native plugin updates and unverified runtime platforms remain guarded.

Refs #409

* fix(omp): guard runtime ownership and localize inventory guidance

Native install recursively replaces its runtime destination, and a missing cache manifest could leave the real runtime link installed after removal. Bind and recheck an absent reviewed destination, and recover removal names only from verified scope-local links.

Localize fixed backend reasons and warnings across dashboard locales while preserving paths and external diagnostics. Restore the reviewed source in the existing reinstall fixture and record real native verification.

* fix(omp): allow forgetting natively removed plugins

A different plugin can retain node_modules after native uninstall deletes this installation. When its cache, registration and known runtime state are absent, forget only the stale Skillshare binding instead of demanding runtime-name proof for unrelated files.

Extend the existing removal recovery test with a failing-before-fix case that verifies the binding is dropped and unrelated native lock bytes and plugin files are preserved. make check passes.
2026-10-05 20:36:39 +08:00
Willie e66529fc76 fix(uninstall): validate linked roots before dry-run output
Dry runs returned before the linked-root guard, so text preview claimed a trash move and JSON listed a root-skill link as removed even though real execution refused it.

Run the existing non-mutating move check before either preview and extend the root-link fixture to verify text and JSON refusals without claiming removal.
2026-10-05 20:36:28 +08:00
Willie cd6e0165dd fix(uninstall): refuse skills at followed link roots
A root skill discovered through a followed source link occupies the link entry itself. Moving that entry to trash silently unlinks every nested skill instead of removing one skill.

Share the root-skill guard across single and batch API handlers and CLI uninstall, report the unlink action clearly, and preserve explicit unlink and nested skill removal.
2026-10-05 20:02:58 +08:00
Willie 98ab6eb832 test(source-links): accept Windows junctions in link assertions
Use the existing link detection and target resolution helpers so correct directory junctions pass the source-link core, API, and CLI tests. Accept the Windows junction kind while retaining symlink coverage and target identity checks without platform skips.
2026-10-05 18:27:12 +08:00
Willie 42ed06e3e4 fix(cli): accept option terminators for link and unlink
Stop parsing mode and command options after -- so valid link names and relative paths beginning with a hyphen can be addressed. Document the unlink form in every command locale and verify that -g and -p after the terminator remain names.
2026-10-05 18:27:12 +08:00
Willie 4ae67cbe52 feat(source-links): add link and unlink commands
follow_source_links only follows a link the user made by hand, so the
guards (source or parent, sync target overlap, link chain) surfaced in
doctor only after the link existed, and Windows users had to pick
between a junction and mklink /D themselves.

link checks the prospective target with Follow.Allow, which shares the
guard rules and reason strings with the walk, before creating a junction
on Windows or an absolute symlink elsewhere. It warns when the target is
not a git checkout and leaves follow_source_links alone unless --enable
is given. unlink moves only the link entry to trash, the same way
uninstall removes a first-level link, so the target is never touched and
trash restore undoes it. internal/sourcelink holds both operations so the
dashboard can call them too.

Refs #419
2026-10-05 18:27:12 +08:00
Willie 3b898f3730 Merge branch 'main' into runkids/274-follow-source-links 2026-10-05 16:11:45 +08:00
Willie eb7b4156bc fix(source-links): report unfollowed links in sync --json and refuse uninstall --group on a link
sync --json showed pruned: 0 and success while a link was unavailable,
so automation could not tell a held-back prune from nothing to prune.
The run's source link warnings now ride along as warnings.

uninstall --group <followed link> moved every skill out of the user's
checkout, while uninstall <link> removes only the link entry. The group
form is refused with a pointer to the by-name forms.
2026-10-05 15:48:45 +08:00
Willie 956db6d122 fix(update): keep --all away from followed checkouts
update --all treated a followed source link as a managed tracked repo:
with --force it ran git reset --hard inside the user's own checkout and
silently discarded their local edits, and an audit-gate rollback would
do the same. --all now skips any tracked repo behind a followed link
with a warning that names the by-name form, which keeps its documented
behavior.
2026-10-05 15:42:01 +08:00
Willie 72ed5ba0fb fix(diff): compare extension agent outputs and apply target filters (#414)
* fix(diff): compare extension agent outputs and apply target filters

#392 taught status and doctor to recognise converted extension outputs,
but diff (CLI) and the Sync tab (/api/diff) still compared the target
directory's *.md files against the source name. A target whose extension
renames the output (codex -> .toml) therefore listed every agent as
pending after a sync.

Both diff paths now use the same manifest-based check as status, and
report tracked outputs whose source is gone as prune items. They also
apply the target's include/exclude and the agents' frontmatter targets,
so an agent that sync would skip is no longer shown as pending.

Refs #391

* chore(release): request 0.24.6

The pending feat commits would make Release Please propose 0.25.0. This
batch is meant to ship as the 0.24.6 patch, so ask for that version.

Release-As: 0.24.6
2026-10-05 15:40:19 +08:00
Willie f9e25ca7a5 fix(audit): report source links the scan did not follow
An audit run with follow_source_links whose linked checkout was missing
finished as a clean success, so a partial scan looked complete. Audit now
warns once per skipped link in human output and adds warnings plus
incomplete: true to JSON output, including when no skills were found.

Also cover project mode: reconcile keeps config and metadata entries of an
unavailable link, and sync -p warns once and keeps target entries.
2026-10-05 15:40:18 +08:00
Willie 8092b234b7 fix(source-links): judge overlap before existence and keep doctor from suggesting prune
A link into a sync target that is not created yet was reported as a
missing target, which marked the inventory incomplete and held back
pruning for a link that would never be followed. The where-it-points
guards now run on the link text before the does-it-exist check, so such
a link is an overlap and the run prunes as usual.

doctor reported the dangling target links behind an unmounted source
link as broken symlinks and suggested `skillshare sync` to prune them,
while sync deliberately keeps them. Those links are now a warning that
names the waiting source link and carries no prune suggestion.
2026-10-05 15:27:42 +08:00
Willie a502e9316b feat(sourcefs): write through followed first-level source links
With follow_source_links on, discovery reports <source>/_dev-skills/foo,
but every write to it was refused: sourcefs rejects a link in the path, and
its source-root os.Root refuses descendants outside the source. Dashboard
edits, nested uninstall, regular skill replacement, and tracked installs
under such a link all failed, although writing through a link the user
placed is the agreed behavior (#410).

sourcefs.Open, Create, MkdirAllIn, and CheckMoveOut take the operation's
optional sourcewalk.Follow. A clean name <link>/<rest> whose first
component the policy follows is handled by a second os.Root opened at the
link's canonical target, with the same per-component link check on <rest>,
so a nested link or ".." still cannot escape the target. Names with ".."
keep the source root, which refuses them through the link. Without a
policy nothing changes.

The link entry itself stays a link: RemoveAll, Rename, and writes on it are
refused, install --track --force still refuses to replace it, and
CheckMoveOut still lets trash rename the link without following it.
MkdirAll on a followed link is a no-op, since it already is that directory;
install --into needs that. A rename across the link boundary is refused.

The install, update, uninstall, and server callers pass the same policy
their discovery uses, through InstallOptions.SourceFollow and the update
context. No raw-write allowance was needed.

Refs #274
2026-10-05 14:42:09 +08:00
Willie a532e77ffb test(doctor): expect the follow_source_links hint in the link message
The doctor message for an unfollowed first-level link now names the config key that enables following, so the integration expectation must carry the hint too.
2026-10-05 14:30:41 +08:00
Willie 89fc4fbdf1 feat(sourcewalk): follow first-level source links when enabled
Users link a checkout directly under the skills source (e.g.
skills/_dev-skills -> ~/code/dev-skills) and expect it to behave like any
other first-level entry. sourcewalk resolved only the root, so these links
were invisible to discovery, sync, update, and the dashboard.

follow_source_links (global and project config, default false) enables a
per-operation sourcewalk.Follow policy. Walk, WalkDir, and ReadDir report a
followed link under its own name as a directory and descend into its
target, so names, RelPath, FlatName, SourcePath, tracked-repo detection, and
metadata keys stay logical. Only entries directly under the source root are
followed. The policy is built from config and passed explicitly to every
discovery caller in the CLI, server, and reconcile, so counts agree; the
server keeps a per-instance snapshot refreshed on config load and save
instead of a process-global flag.

A link is skipped, with one warning per operation, when its target is
missing or unreadable, is the source or a parent of it, or overlaps an
active skills target. A missing or unreadable target marks the operation
incomplete: sync then skips pruning and copy-orphan deletion, and reconcile
keeps metadata, so an unmounted drive or a re-clone in progress cannot
erase target entries or install records. A link to an existing file is not
a directory link and stays an ordinary entry; treating it as incomplete
would disable pruning for anyone sharing a file such as .skillignore.

doctor keeps raw link identity: with the key off it still says the link is
not followed and names follow_source_links: true; with it on it reports
followed links and skipped links with the reason.

The raw-walk allowlist gains entries for the policy path in Walk and
WalkDir: the source root walk with a policy, and the walk of a validated
link target whose paths are remapped to the logical link path. Both live
in sourcewalk itself, the package that owns source traversal.

Refs #274
2026-10-05 14:26:09 +08:00
Willie 47a98b81ff fix(status): expect only the agents a target would sync
status counted every discovered agent against each target, while sync
agents writes only the agents left after .agentignore, the target's
include/exclude and the frontmatter targets field. A target that leaves
some agents out was reported as N/M with drift after a complete sync,
in the table and in --json, in global and project mode. doctor and the
dashboard already filtered per target.

status now applies the same three filters before counting and expecting.
The table reports an invalid include/exclude as an error, like it does
for skills; --json falls back to the active agents.

Refs #395
2026-10-04 22:21:00 +08:00
Willie 51a3be1ca6 fix(status): report drift when an extension agent's source changed
Counting an extension output as synced from its output hash alone missed
the common case: the user edits the agent in the source and runs status
to see whether a sync is due. The output still matched the manifest, so
status, doctor and the dashboard reported no drift although the next
sync would regenerate it.

Sync now records a SHA-256 of the source each output was converted from
in the manifest (sources), and status counts an output only when both
the output and that fingerprint are current. Outputs recorded by earlier
releases have no fingerprint and count as current until the next sync
records one, so upgrading does not show the 0/N the report was about.

Refs #391, Codex review on #392.
2026-10-04 19:49:29 +08:00
Willie f403ea5a6f fix(status): count extension-converted agents as synced
An agent target with an extension writes each agent under output_ext
(reviewer.toml) with transformed content, and the manifest tracks that
output name. status, doctor and the dashboard's Sync tab counted synced
agents by looking for the source name and comparing with the source
content, so every converted agent was missing: status showed 0/N with
drift, doctor said the same, and the Sync tab kept them pending after a
successful sync.

For an extension target, count an agent as synced when the manifest
tracks an output with the same stem and that file still matches its
recorded hash. Matching by stem avoids loading the extension spec in
status paths. doctor keeps reporting leftover broken links for these
targets.

Refs #391
2026-10-04 19:33:50 +08:00
Willie 8c60f51b9f feat(source): add the sourcewalk walker and sourcefs write boundary for proposal 274 (#385)
Implements rollout steps 1 and 2 of proposals/274-skillfollow.md. Nothing follows a link or parses .skillfollow yet.

Step 1: doctor reports first-level symlinks and junctions in the skills source as info, in text and JSON, without following them.

Step 2a: internal/sourcewalk owns traversal of the skills source. Thirteen raw walks move onto it with identical behavior, and a go/parser ratchet fails the tests on any new unclassified walk.

Step 2b: internal/sourcefs wraps an os.Root opened at the skills source and refuses any write whose path has a link component. The six write sites the proposal found in review move onto it, including the cross-device copy fallback, and a second ratchet covers raw os write calls. The only behavior change is that an explicit path crossing a link inside the source is refused instead of written through.

Refs #274

Release-As: 0.24.3
2026-10-04 15:15:43 +08:00
Willie 9ac9682dcc perf(memory): tell agents to prefer shared notes over tool memory
The generated guidance block defined "what is worth saving" the same way
agents' built-in memory does, but never said how the two relate, so agents
with their own memory kept saving there and left the shared folder alone.
The active rule now says to save here instead of tool-specific memory,
gives a note format and an INDEX.md line shape, and both modes read
INDEX.md at the start of each task instead of "when relevant". The
passive rule asks agents to point out facts worth keeping so the user
can ask for them to be saved.

Dashboard mode descriptions, the CLI help, and the website pages are
updated to match. Test fixtures that anchored on "Read only" follow the
new wording.
2026-10-04 14:45:38 +08:00
Willie b4296e3cff feat(memory): let each agent choose passive or active note updates (#377)
* feat(memory): let each agent choose passive or active note updates

Shared memory guidance only told agents to update notes when asked, so
anything worth keeping was lost unless the user remembered to request it.
Some users want agents to maintain the notes themselves, but because the
notes are shared across every tool, blind writes would fill them with
noise.

Add an active mode whose guidance limits saves to facts that will matter
in later sessions, asks the agent to propose a note when unsure, to update
an existing note instead of duplicating it, and to report what it saved.
Passive stays the default and its block is byte-for-byte unchanged, so
existing connections remain configured.

The mode is recorded in each block's marker, so agents can differ. Agents
that read the same file share one block; the dashboard switches them
together and the API rejects conflicting modes for one file. The CLI
prints either block with `extras memory instructions --update-mode`
(`--mode` already means an extra's sync mode).

* feat(memory): review guidance changes as a unified diff

The side-by-side Before/After panes made users scan two windows to find
the one block that changed. A single Git-style diff shows only the
removed and added lines in context, with removals listed first.

* fix(memory): flag targets that read guidance in both update modes

A tool that imports several shared instruction files can read a passive
block in one and an active block in another. The status showed only the
best-ranked block, so the target looked configured in one mode while the
agent still received the other, and switching modes rewrote just one
file. Report such a target as needing attention instead, so the user
sets the tools on those files to one mode.

* fix(memory): refuse mode switches that would split a target's guidance

A target can read intact blocks in more than one file, such as its own
file and an imported shared file. Switching its mode rewrote only the
file the status pointed at, leaving the other block in the old mode and
creating the contradiction the mixed_modes state reports. Skip such a
switch as multiple_blocks so the user removes the extra block first.

Document guidance modes, mixed_modes and multiple_blocks in the
architecture topic, which still described a single implicit mode.
2026-10-04 10:23:47 +08:00
Willie dd6991ca3c fix(push): refuse unpushed config.yaml history at root scope (#366) (#372)
* fix(push): reject unpushed config history at root scope

Untracking config.yaml at the tip still uploads its contents from earlier local commits. Refuse root-scope pushes before mutation and before upload when unpublished history adds or changes the file or its directory tree, without rewriting history.

Share the check across the CLI and dashboard, return unpushed_config_history, and allow removals and config inherited unchanged from upstream merges. Cover first pushes, dry runs, push --pull, directory paths, renames, and merge commits, and document manual repair in all website locales.

Fixes #366

* fix(push): check exactly what push sends for config history

Review of the config history guard found two range problems:

- The guard checked @{u}..HEAD, but a same-named upstream was pushed with
  a bare `git push`. With remote.pushDefault or push.default=matching that
  could send config-bearing commits to another remote, or other branches,
  that the guard never looked at. PushArgs now always names the upstream
  remote and branch (read from branch config, so names with "/" survive),
  so the checked range and the pushed range are the same.
- Without an upstream every commit on HEAD counted as unpushed, so a
  config.yaml already published on origin followed by the
  "Keep config.yaml out of version control" commit was refused. The guard
  now checks HEAD --not --remotes=<push remote>.

The refusal message now suggests rebasing from just before the oldest
offending commit instead of @{u} or --root, which would rewrite commits
the remote already has.

* fix(push): check config history against the remote's live refs

Local remote-tracking refs can be stale. When another machine force-pushes
the remote to drop a leaked config.yaml commit, this clone's
refs/remotes/<remote>/... still contains it, so the guard treated the leak
as already published and HEAD:main, still a fast-forward, would publish it
again.

The guard now excludes only the objects the push remote's refs point to
right now (git ls-remote), limited to objects this clone has. It falls back
to local remote-tracking refs only when the remote cannot be reached, in
which case the push fails anyway. Revisions go to git log on stdin as
"^<id>" lines because git before 2.42 rejects "--not" there.

* fix(push): screen config history against every push URL, without replace refs

Two more ways the screened history could differ from what git push uploads:

- `git ls-remote <remote>` queries the fetch URL, but git push sends to every
  remote.<name>.pushurl. A commit already on the fetch URL could still be
  new to a push URL. The guard now lists `git remote get-url --push --all`
  and excludes only objects that every push URL has.
- With a refs/replace/* entry for an offending commit, git log read the
  sanitized replacement while the push pack carries the original. The log
  and the rebase-base lookup now run with --no-replace-objects.

* fix(push): let pushes the remote will reject reach the pull step

Our own review of the config history guard found two problems:

- When another machine had pushed, the remote's tip was unknown locally, so
  no live ref survived the "objects this clone has" filter and the whole
  history counted as unpushed. A config.yaml published long ago, followed by
  the stop-tracking commit, then blocked `push --pull` before it could pull,
  and replaced the plain push's ErrPushRejected hint. A push URL whose
  destination branch tip this clone lacks rejects the push as
  non-fast-forward and receives nothing, so it is now left out, and the
  check passes when every push URL is like that. `push --pull` re-runs the
  check after pulling, with the tip known.
- PushRemote read branch.<b>.remote even when @{u} no longer resolved, while
  PushArgs then pushed to origin. pushTarget now follows the same
  HasUpstream rule as PushArgs.
2026-10-04 03:09:17 +08:00
Willie 475b176914 fix(pull): keep the local config.yaml when a later pull brings a tracked copy (#362)
* fix(pull): keep the local config.yaml when a later pull brings a tracked copy

#357 refuses a root-scope first pull when the remote tracks config.yaml, but once upstream is set, a pull that merges in a remote-tracked copy still replaced this machine's config: Git treats the ignored file as expendable. Snapshot config.yaml around later pulls (CLI pull, push --pull, and the dashboard), restore it, and warn so the user can untrack it with push.

Refs #353

* fix(git): untrack config.yaml after push --pull and warn on identical copies

push --pull ran the root safety sweep before integrating the remote, so a
pull that brought in a remote-tracked config.yaml was pushed back as-is:
the remote kept tracking it and the next first pull elsewhere was refused.
Re-run EnsureConfigUntracked after the pull and commit the removal before
pushing.

The warning also depended on the pull changing the local bytes, so a
tracked copy identical to this machine's went unreported. KeepLocalConfig's
restore now reports whether the repo tracks config.yaml after the pull.

* fix(git): only restore config.yaml when the pull could have touched it

KeepLocalConfig wrote its snapshot back after every root-scope pull, so an
edit made to an untracked config.yaml while the pull ran was silently
replaced by the stale copy. Git only rewrites config.yaml when it is
tracked before or after the pull, so restore only in that case.

* fix(git): restore config.yaml only when the pull changed its index entry

Gating the restore on "tracked before or after" still overwrote edits made
during a pull when config.yaml was already tracked and the pull left it
alone. Compare config.yaml's index entry (mode, blob, stage) before and
after the pull instead; a merge, reset or conflict that touches the file
changes it, and anything else keeps the current file.

* fix(git): warn about a tracked config.yaml only when the upstream tracks it

The warning followed the local index, so a config.yaml tracked only by a
local, unpushed commit was reported as tracked by the remote, and the
suggested push would publish it in history. Check the upstream tree with
RemoteTracksConfig("@{u}") instead.

* fix(git): restore config.yaml after a pull that aborts its merge

A merge that brought in a tracked config.yaml and conflicted elsewhere was
aborted, which deleted the file again and returned the index entry to its
pre-pull state, so restore saw no change and the local config was lost.
Also restore when the snapshotted file is missing after the pull.

* fix(git): keep config.yaml ignored when a later rule negates it

EnsureConfigUntracked only checked that a "config.yaml" line existed, so a
pulled .gitignore with a later "!config.yaml" left the file unignored and
the following `git add -A` tracked this machine's copy again. Check the
effective ignore state and append an overriding entry when needed.

* fix: commit post-pull .gitignore repairs and keep config.yaml's mode

push --pull committed only when EnsureConfigUntracked removed an index
entry, so a pulled .gitignore that stopped ignoring config.yaml was
repaired locally but never pushed, leaving the tree dirty. Commit whenever
the safety step leaves changes.

KeepLocalConfig skipped the restore when the bytes matched, so a checkout
that rewrote config.yaml with the default mode dropped this machine's
restrictive permissions. Compare the mode too, and chmod after writing
since WriteFile's mode is masked by umask.

* fix(git): restore config.yaml over a tracked config.yaml directory

A pull that checks out a tracked config.yaml/ tree replaces the ignored
local file with a directory, which os.Remove cannot delete, so restore
failed and the local config stayed lost. Remove a directory there before
writing the snapshot back; its contents are in Git. EnsureConfigUntracked
now untracks recursively so push can remove that tree too.

* fix: keep post-checkout config edits and preview .gitignore repairs

Restore wrote the snapshot back whenever the pull changed config.yaml's
index entry, even when a newer copy had been written after Git checked it
out. Skip the restore when the file on disk no longer matches the blob the
pull checked out.

A dry-run push or commit only previewed untracking config.yaml, so when a
later "!config.yaml" rule un-ignored it the preview listed config.yaml to
commit while a real run repairs .gitignore instead. Preview that repair.

* fix: scope the post-pull safety commit and keep post-checkout symlinks

The post-pull safety step in push --pull ran `git add -A`, so an edit made
by another process while the pull ran was committed and pushed under the
config-safety message. Stage only .gitignore on top of the config.yaml
removal and commit the index as staged.

The post-checkout edit check skipped symlinks, so a dotfiles manager that
replaced config.yaml with a newer link after the checkout had it replaced
by the stale snapshot. Treat a link where Git checked out a file, or a
different link target, as a newer local edit.

* fix: preview explicit config.yaml entries and refuse other .gitignore edits

A dry run previewed the .gitignore repair only when config.yaml was not
ignored at all, but a real run also appends the explicit entry when a
broader rule such as *.yaml ignores it. Preview whenever the entry is
missing too.

The post-pull safety commit staged the whole .gitignore, so an edit made
by another process while the pull ran was pushed with it. Refuse unless
the file only gained config.yaml lines over its staged version.

* fix: report config.yaml restore failures instead of dropping them

integrateRemote declared the KeepLocalConfig result with :=, shadowing the
named err that the deferred restore joins its failure into, so a failed
restore was swallowed and pull or push --pull went on to sync targets with
another machine's config.yaml. The dashboard pull replaced the pull's own
error with a bare 500 on a restore failure, losing the conflict response
and the ops log entry; join the errors so both are reported.
2026-10-04 01:58:30 +08:00
Willie f76af3f0f6 test(plugin): resolve the sandbox home in the Pi npm test (#364)
Native config writes refuse any symlinked ancestor, and macOS temp
directories live under /var, a symlink to /private/var. The test pointed
PI_CODING_AGENT_DIR into the unresolved sandbox home, so the macOS
integration job blocked the change and failed on main. Resolve the home
first, as the hooks test already does.
2026-10-04 01:13:56 +08:00
Willie 3ba95ef49e feat(plugin): add npm packages from pi.dev through Pi (#363)
* chore(design): export dashboard tokens for design tools

Mockup tools read a design system's tokens.json rather than our CSS, so
designs drifted from the dashboard's Clean theme. The script derives the
tokens from components.css, and CI runs its test so a new color
variable without a usage note fails instead of silently going missing.

* feat(plugin): add npm packages from pi.dev through Pi

Pi packages on pi.dev are npm packages, which Plugins could not add:
it only took local or Git sources, so users had to install them with Pi
and import them. `plugin add npm:<package>` now hands them to
`pi install` for Pi targets and records the binding like an import.

Skillshare can't fetch or review an npm package, and Pi runs its install
scripts, so the preview says so, discover refuses npm sources, and only
targets that run pi itself take them. Pi keeps one entry per package
name, so another version replaces it in place and keeps its extension
filters; Skillshare records the new entry so a reinstall restores them,
and refuses a package another Skillshare package already manages, since
removing one would remove both. Pinned versions are skipped on update
because pi update keeps them.

The dashboard accepts npm sources, a pasted `pi install` command or a
pi.dev address, lists only Pi targets, and the Pi Extensions tab links
to adding a package with that target selected. Pi packages show Pi's
mark in the plugin list.

A blocked change now keeps its own reason when the Agent's CLI is also
missing, and the message key always matches the message the CLI prints.

* fix(plugin): keep npm package names and accept pi by path

The default name of an npm package went through the file-path rule, which
cut its last dot segment as an extension: npm:foo.bar was recorded as foo,
and adding npm:foo.bar@1.2.3 later derived foo.bar and was refused as
another package managing the same Pi entry. npm names are now taken from
the package spec.

A Pi account whose CLI is an absolute path to pi was treated as running
another executable, so it could not take npm packages. It now uses the
same path- and platform-aware check as the Pi Extensions tab.

* docs(wiki): add a pull-requests topic for Codex review rounds

Handling Codex's review was only known from practice: confirm each
finding, fix it with a failing test first, reply in its thread, ask for
another review, and follow up until it answers. Writing it down keeps
agents from resolving threads, amending pushed commits, or stopping at
the 👀 that only means the review is still running.

* feat(ui): give the Extras AGENTS.md and Memory tabs one layout

The two tabs shared a page but looked like different products: the header
shifted between them, notes sat in a joined box with an icon toolbar, and
agents showed as an icon strip in a side card. Switching tabs felt jarring.

Memory now follows the AGENTS.md layout: a sticky list column with search,
the note's name with Edit and a More actions menu, the same preview box,
and Use with agents as one row per tool under the note. Long notes open
collapsed like shared files do. The AGENTS.md list gains a search box and
the same sticky column. The Memory tab label stays English in every locale,
like AGENTS.md. Docs and screenshots follow the new layout.

* chore(docs): add a script that serves the Memory screenshot homes

The Memory how-to screenshots came from demo homes that were never
recorded, so reshooting them meant rebuilding the data by hand. The script
rebuilds the fresh and the populated home inside the devcontainer, serves
each behind the Vite dev server, and hands the port back to air afterwards.

* fix(plugin): refuse npm sources Pi can't resolve

validNpmSource accepted specs such as npm:foo/bar, npm:foo@ and names
containing "..", which Pi's own parser rejects. They passed preview, and
Apply recorded a pending binding before pi install failed, leaving an
entry every later sync retried. Validate with the same grammar the Pi
settings reader uses, which also covers a scope without a package name.

* fix(ui): show the default name when adding an npm package to Pi

The name field for an npm source was blank, so it looked like a name was
missing. Leaving it empty records the package name without its version or
scope, as the git flow already shows; the field now shows that name as its
placeholder.
2026-10-04 00:43:43 +08:00
adam maguedandWillie 7fccd6ffa5 fix(pull): refuse root-scope first pull when remote tracks config.yaml (#357)
When git_root is configured as root, config.yaml is machine-specific and kept out of version control by .gitignore. If a remote repository tracks config.yaml, a first pull silently overwrites the local config.yaml during git merge or git reset.

Detect whether the remote tracks config.yaml at root scope before integrating changes in FirstPull and init remote setup. If so, refuse the operation and report guidance to untrack config.yaml on the remote via push before pulling.

Co-authored-by: Willie <egg358358@gmail.com>
2026-10-03 23:50:35 +08:00
Willie 539f1d0f03 fix(pi): recover Windows stale locks and protect private registrations (#361)
* test(pi): reproduce Windows private registration access failures

Use NTFS DACL assertions instead of POSIX mode bits and verify that broadly accessible existing records cannot be read or silently reused. Keep the Unix mode assertion unchanged.

* fix(pi): reclaim Windows locks and protect private registration state

Anchor stale directories with delete sharing and delete the verified empty Windows directory by handle, keeping inode, lease and racing-owner guards. Create new registration directories with protected owner/SYSTEM DACLs and validate existing owner/access rules without changing their ACLs. Leave Unix permissions and directory-only removal unchanged.

* test(pi): capture Windows lock identity before path replacement

Windows Lstat lazily fetches file IDs by pathname. Snapshot the stale fixture through its anchored handle and add a matching-mtime replacement regression before first verification or release.

* fix(pi): snapshot Windows lock IDs and accept privileged default owners

Capture acquired lock identity through a handle before Windows can lazily resolve a replaced pathname. Elevated tokens can default file ownership to Administrators; permit privileged owners only alongside a verified current-user-only DACL. Refuse unsafe existing registration directories without modifying their ACLs or writing records.

* docs(pi): record Windows acceptance and isolated regression workflow

Document private Windows ACL requirements and refusal without permission rewrites. Add pinned devcontainer kit packaging, token-isolated full/basic acceptance, preserved UTF-8 logs and a reproducible UTM runbook. Preserve the original Unix permission assertion.

* test(backup): isolate explicit agent backup from sync snapshots

Sync automatically backs up target-local agents. Creating the fixture before sync makes the exact-one-backup assertion depend on both commands sharing a timestamp second. Create it after successful sync so the explicit backup is the only operation that snapshots it, without weakening the assertion.
2026-10-03 23:19:34 +08:00
Willie d78e7ce09a feat(cli): redesign init and give every command one output style (#360)
* fix(install): omit the empty audit section in dry-run

A dry-run skips the security audit, yet the single-skill path still printed
the "Audit Findings" heading with nothing under it, which reads as if the
audit ran and found an empty result.

* feat(init): ask first, confirm a summary, then write

The old init asked nine questions in three input styles and wrote files as
it went, so cancelling midway left a half-built setup, and detection created
tool folders even under --dry-run. Connecting a second machine to an
existing repo was only offered at the very end, and the built-in skill
defaulted to "no", which read as "not recommended" (#200).

init now collects every answer into a plan, shows a summary with Yes /
Change settings / Cancel, and writes nothing until confirmed. Esc cancels
cleanly. A "Connect my existing skillshare repo" path checks the repo
before writing, detects its layout, prefers the repo version of same-name
skills and keeps local-only ones for the next push. The built-in skill is
installed by default, and the first sync runs at the end so the tools work
right away. Prompts use one inline huh-based component with theme colors.

Without a terminal (CI, scripts, AI agents) init now applies the same
defaults as pressing Enter and prints each decision with the flag that
changes it, instead of silently doing less; init -p and --discover add the
detected tools. Output that is not a terminal is plain text.

BREAKING CHANGE: non-interactive `init` now enables git, installs the
built-in skill and selects every detected tool; pass --no-git, --no-skill,
--no-targets or --no-copy for the previous minimal setup. Piped or
redirected output no longer contains ANSI colors; use --json where offered
for machine-readable output.

Refs #200

* docs(init): show the new init flow and the connect-a-repo path

The getting-started page still described the old four-question wizard with
a screenshot of it, the second-machine guide never mentioned the new
"Connect my existing skillshare repo" choice, and the recipe told users to
run pull and sync after init even though init now does both.

Replace the screenshot with ones from the new flow, describe the questions,
summary and cancel behavior, and explain what connecting a repo does before
and after confirmation. The devcontainer example gains --no-git because
init without a terminal now turns git on by default.

Refs #200

* feat(devcontainer): seed fake tools and a repo with ssenv --seed

Trying init by hand needed several manual steps to fake installed AI tools
and a skills repo to connect. `ssenv create --seed` (and `ssnew --seed`)
now creates claude/codex/cursor directories with a few skills and a local
bare repo at ~/remote/skills.git, so both init paths can be tried at once.

Refs #200

* fix(ui): skip the logo on 16-color terminals and hide the cursor while it plays

`docker exec -it` sets only TERM=xterm, which reads as 16 colors: the logo
was mapped onto the terminal's own palette and its fade-in collapsed into a
few flickering cells. Below 256 colors the banner now prints only its text,
and the devcontainer sets COLORTERM=truecolor.

The visible cursor also jumped along the logo's left edge while frames
redrew in place. It is hidden during the animation and restored on exit,
including Ctrl+C.

Refs #200

* style(ui): color the question mark in prompts

The uncolored "?" before each question read like a stray character. It
now uses the accent color of the "›" cursor.

Refs #200

* fix(init): name the root git scope by what it versions

"everything" suggested plugins, MCP servers and hooks were versioned too,
but they live in config.yaml, which the root scope gitignores. The option
and summary now say "skills, agents, extras" and note that those settings
stay in each machine's config.yaml.

Refs #200

* feat(init): show --discover in the new init style

--discover still printed the old header, "agent(s)" counts and "+ name"
lists next to the new prompt. Global and project discover now share
init's answered rows (Tools, Config) and a Next line, and call the
targets "AI tools" as init's banner does.

Refs #200

* feat(init): replace the pixel logo with a SKILLSHARE wordmark

The half-block logo looked stretched on terminals with tall line height
and turned to mush when shrunk to fit. init now opens with a three-line
pixel wordmark, SKILL in blue and SHARE in gold, that lights up from left
to right, followed by the version and tagline on one dim line. The
generated logo data and its generator script are gone.

Without a terminal, without color, below 256 colors or on a narrow
terminal, the banner still prints a plain "skillshare vX" line.

Refs #200

* refactor(init): label the chosen tools "Targets"

init called them "Tools" while the rest of the CLI, the config file and
the `target` command call them targets. The answered rows and the summary
now say "Targets", the same label the other commands will use. Sentences
that describe what was detected ("Found 3 AI tools") stay as they are.

The docs tables, the First Sync steps and both init screenshots are
updated to match.

Refs #200

* feat(cli): ask yes/no questions with the inline prompt

uninstall, trash empty, extras remove, target add, collect, update,
upgrade and audit each printed their own "[y/N]:" line and parsed the
answer by hand, while init already used the inline Yes/No prompt. They
now share it: esc answers no, and a declined question prints one dim
"Cancelled. Nothing was …" line, as init's cancel now does too.

Without a terminal the prompt reads one line from stdin instead, so
piped answers such as `echo y | skillshare uninstall x` keep working.
It reads byte by byte so a second question still gets its own answer.

Refs #200

* feat(sync): report each target on one aligned row

Sync printed a ruled header, a "Backing up" block and a "name: merged
(43 linked, 0 local, 0 updated, 0 pruned)" line per target, so the
counts that changed were hard to find among zeros. It now follows the
init style: one row per target with only the non-zero counts, a single
Backup row, and a closing line with the duration. `sync --all` names
its Skills, Agents and Extras blocks and no longer explains how to set
up extras when none are configured.

The non-TTY text changes with it; `--json` stays the stable interface
for scripts. Tests match rows through the new AssertRowContains helper.

Refs #200

* feat(status): show targets as one table

Status printed each target as a header with indented skills and agents
lines full of bracketed modes and zero counts, and the audit policy as
four lines of uppercase labels. Comparing targets meant reading every
block. Each target is now one row with a skills and an agents column,
a single line names the modes in use, and audit and version take one
line each. Global and project mode share the renderer, so project
paths read relative to the project root.

Refs #200

* feat(install): report each step on one aligned row

Install printed a logo, a box-drawn tree, "Next Steps" bullets and a
separate "Audit Findings" block, so the result was hard to find. It now
uses the same rows as sync and status: Source, Audit and Install with
the destination folded to ~, then a closing line and a Next block.

Spinners write through a gate that Stop closes before clearing the line.
pterm's render goroutine checks IsActive and then writes, so a fast step
such as install --dry-run left "Copying files..." stuck in front of the
next line.

Refs #200

* feat(uninstall): confirm against one line and point at undo

Uninstall opened with a boxed header and Name/Path lines, and closed
with the trash folder path and a "Next Steps" bullet. Now one line names
the target, its folder and what it holds; the result is a single
"Uninstall x → trash, kept 7 days" row; and Next offers sync, the trash
restore command that undoes it, and the reinstall command when the
skill came from a remote source.

A batch run with --force or --dry-run no longer lists the targets twice,
since there is no question for the list to support. Mixed batches are
counted as items, not targets, which already names the AI tools.

Refs #200

* feat(update): report each skill or repo on one Update row

Update opened with an "Updating" header, the full source path and a
"0 tracked repo(s), 1 skill(s)" line, then printed a Changes box and two
success lines before "Next Steps". A single update is now an Update row
(source or commit count, time), the commits it pulled in dim text, the
Audit row, and Next. A batch ends with the audit result and one closing
line that leaves out zero counts.

Progress bars now clear themselves when they stop, like spinners, so a
finished batch no longer leaves "[3/3] ... 100% Done" behind.

Refs #200

* feat(check): list what needs attention and close with one line

Check opened with a header, the full source path and an item count,
counted skills with updates without naming them, and ended with two
"Run ..." hints. It now prints one row per repo, skill or agent that
needs attention (behind, update available, stale, failed), a closing
line that leaves out zero counts, and the update command under Next. A
single-name check is one row, like update.

Refs #200

* feat(push,pull): report commit, push and pull as aligned rows

Push and pull opened with a header and closed with "Push complete" or
"Pull complete", so they never said what they committed, where they
pushed, or what a pull brought in. Push now prints a Commit row (file
count and message) and a Push row naming the upstream branch, never the
remote URL since it can carry credentials. Pull prints a Pull row with
the commit and file counts and the commits themselves, as update does
for a tracked repository, before the sync output. Dry runs list the
changed files as ~ + - instead of raw porcelain codes, and error hints
are dim notes instead of info lines.

Refs #200

* feat(upgrade): report the CLI and the skill on one row each

Upgrade opened with the logo and printed a tree of steps per part
(current version, latest version, status) ending in "Already up to date
✓". Each part is now one row, such as "✓ CLI  v0.22.0 → v0.23.0 · 3.1s"
or "✓ Skill  v0.23.0, already up to date", with spinners that clear.
Sync is offered under Next only when the skill changed, and the star
request is a dim line at the end.

Refs #200

* feat(new): report the created skill on one row

New printed a "New Skill Created" header, the full path, and numbered
"Next steps" that told the user to edit the file and run sync. It now
prints "✓ Created  ~/…/SKILL.md", the folders a pattern added, and sync
under Next. A dry run shows the same row, a Preview section with the
template, and the dry-run line.

Refs #200

* feat(trash): list items and restores as aligned rows

The plain trash list printed each item as an info line with its size and
age in parentheses under a "Trash" header, and restore printed two
success/info lines plus a "Next Steps" label. The list is now rows of
name, size and age with one closing count, and restore is one row
naming where the item went back, with sync under Next. An item trashed
under a minute ago reads "just now" instead of "just now ago".

Refs #200

* feat(collect): list local items and results as aligned rows

Collect opened with a header, listed local items as "[target] path"
list items under a second header, then printed a third header before
the per-item results and a metric summary line. It now lists each local
item as a row (name, target, path), asks the inline confirm, prints one
row per result and closes with "Collected N skills" plus sync and,
when the source is a git repository, commit under Next.

Refs #200

* feat(hooks): show hooks and their changes as aligned rows

Hooks printed "Hooks source: <path>" as an info line, statuses through
the fixed-width Status helper, and results as sentences ending in
"Run 'skillshare hooks sync' when ready." Hooks and Agent file changes
are now rows, with conflicts marked as failures and inactive outputs as
warnings, saves and applies close with one line, and the next command
sits under Next. The list's "Pending sync" section no longer shows
entries that are already in place.

Refs #200

* feat(mcp): show servers, changes and checks as aligned rows

MCP printed "MCP source: <path>" as an info line, plan changes through
the fixed-width Status helper, results as sentences that named the next
command inline, and check summaries as "2 server(s) checked: 1
error(s)". Changes and parked servers are now rows with conflicts
marked as failures, saves and applies close with one line and Next, and
check marks each server with its findings beneath and closes with
"Checked 2 servers: 1 error". The Status helper had no callers left and
is removed.

Refs #200

* feat(plugin): show packages, plans and results as aligned rows

Plugin text output was raw "name · target · id · state" lines with
"Warning:", "Includes:" and "Revision:" prefixes, and the list walked a
map so packages came out in a different order each run. Packages,
discovered candidates, planned changes and results are now rows sorted
by name, marked by state (registered, not installed, pending, blocked,
failed), with details as dim notes and the dry-run line after a
preview. JSON output is unchanged.

Refs #200

* feat(audit): replace header and summary boxes with rows

Audit framed its context in a header box (mode, path, block rule,
policy), numbered batch results "[1/2]", printed per-skill verdicts as
"→ Aggregate risk / Block decision" lines, and closed with a Summary box
that listed every count including zeros. The context is now one title
line with a dim note, flagged skills are aligned rows with their highest
severity, the summary leaves out zero counts, and the run closes with
one line saying whether anything was blocked and by which severity,
followed by the command to see the first blocked skill's findings.

The progress bar no longer writes its final frame when stdout is not a
terminal; that frame had no newline and ran into the next line.

Refs #200

* feat(audit): show audit rules as aligned rows

The rules list was a fixed-width table framed by 90-character dividers
with a "(N rules: N enabled, N disabled)" footer, and rule changes were
sentences like "Set severity for X → HIGH". Rules are now rows keyed by
ID with a colored severity and the pattern dimmed, disabled rules dim
as a whole, the list closes with a count that names disabled rules only
when there are some, and each change is one row.

Refs #200

* feat(search): drop the logo and tree, show results as rows

Search opened with the ASCII logo, drew its progress as a ├─/└─ tree,
listed results under a column header with dividers, and ended installs
with "Run 'skillshare sync' to distribute to all targets". Results are
now aligned rows with the source, stars or risk, and the description
and tags beneath; a single install reports the same Install row as
`skillshare install`; batch installs list only failures before one Done
line; and sync and audit sit under Next. Login and rate-limit problems
name the command that fixes them. Non-terminal output no longer prints
a "name: installing..." line per skill.

Refs #200

* feat(backup): show backups and restores as rows with one closing line

Backup and restore opened with headers, announced dry runs with a
warning before doing anything, summarized with an operation box, and
followed a backup with a "Recent backups" list. Targets now back up in
name order as rows with folded paths, runs close with one Done line
(and Next restore), listings are rows keyed by timestamp or path with a
total, and dry runs end with the shared dry-run line. Cleanup reports
how many backups remain instead of a separate "Current" line.

Refs #200

* feat(help): one line per command, subcommands in each command's help

The top-level help opened with the ASCII logo and listed about 70 lines
across ten uppercase sections plus 15 examples, so it did not fit a
screen, yet it left out analyze and the common options. It now lists
each command once in six groups with the common options, and points to
`<command> --help` for subcommands. The mcp, plugin, hooks and extras
pages use the same layout and keep all their content.

A mistyped command prints one error with the closest command instead
of the whole help, a lone option such as --no-tui shows the help, and
`skillshare version` prints one line.

Refs #200

* feat(extras): show extras and their changes as rows with Next

Extras printed a header, "→ name" lines with ad-hoc colored marks
(including ✗ for targets that were only waiting for a sync), and ended
each change with a sentence naming the next command. The list now shows
each extra as a bold name over aligned target rows (! for drift or not
synced) and closes with a count and, when something is pending, Next
sync extras. Init, collect, add/remove target, mode and source changes
close with one Done line, details become notes, and dry-run collect
ends with the shared dry-run line.

Refs #200

* feat(target): show targets as sections and rows with Next

Target info and the plain list printed "Key:" lines at two indent
levels, with "(none)" for every unset filter, and remove printed
"Backing up" / "Unlinking" headers with one line per step. Info now
shows the name over Skills and Agents sections with aligned rows, the
list gives each target one row per resource with details dimmed, and
empty filters are left out. Remove reports one row per target (plus
its backups) and closes with a Done count; dry runs say what would
happen per target and end with the shared dry-run line. Skills on/off,
add and mode/naming/filter changes close with Done and Next.

Refs #200

* feat(doctor): report checks as aligned rows and end with Next

Doctor opened with the ASCII logo, split its report under divider
headers, mixed ✓/→/! lines with full paths and "(s)" counts, listed
targets in random map order, and ended with a bare warning count that
left the fix to scattered "Run:" hints. Checks now print as aligned
rows under bold sections (Environment, Targets, Extras, MCP/hooks/
plugins, Storage, Version), paths fold to ~, targets are sorted with
one row per resource, and the report closes with a Done line plus up
to three Next commands collected from the checks (sync, sync agents,
sync hooks, upgrade, ...). The --json output is unchanged.

Refs #200

* feat(hub): drop the logo and report hub changes with Done

hub index opened with the ASCII logo and a tree of spinner steps, then
repeated the result as "→ Mode / Skills / Output" lines with the full
path. It now shows one spinner that clears itself and a Done line
naming the folded output path, skill count and duration, with the
index contents as a note. hub list shows aligned rows with the default
and built-in hubs marked in dim text instead of a "*" column, and the
add, remove and default commands close with a Done line.

Refs #200

* test(install): match the shared dry-run line in online tests

The online-tagged install tests still looked for the old "dry-run"
banner, which install no longer prints since the restyle. They are
skipped by default, so the mismatch only showed under -tags online.

Refs #200

* fix(target): align remove rows on target names only

The label column was sized for "<name>-agents" so an agents backup
could fit, which left a wide gap on every row when no agents were
backed up. The agents backup row now uses the target name too and says
"agents backed up", so the column fits the names.

Refs #200

* fix(update): list each item in a batch dry run

A batch dry run only said "Would update N skills". Off a terminal the
names leaked out through the progress bar's title lines, but on a
terminal those lines clear, so the dry run never said which skills or
repos it would touch. It now prints one Update row per item, as the
single-item dry run does.

Refs #200

* fix(ui): keep the progress bar silent off a terminal

Off a terminal the bar printed "Title (0/N)", every title change and
every phase header as permanent lines, which a terminal shows only as
a frame that clears itself. Audit left a stale "Scanning skills (0/7)"
line, and scripts saw different content than a terminal. The bar now
prints nothing off a terminal, like the spinner.

Refs #200

* fix(audit): avoid a double blank line before Summary

The audit header ends with a blank line and Summary added its own, so
a clean scan with no result rows showed two blank lines in a row. The
Summary title no longer adds one; the result rows add it when present.

Refs #200

* fix(mcp): drop the blank line above "Saved the MCP source"

A save with no backup or migration notes printed a blank line with
nothing above it, so mcp add and mcp edit opened on an empty line.

Refs #200

* fix(doctor): stop repeating the extra's name in its findings

Extras findings printed "rules  rules → /home/me/.claude/rules: 1
file(s) out of sync": the row label already names the extra, the path
was not folded to ~, and the count used "file(s)". The rows now start
at the arrow, fold the path and pluralize the count. JSON details keep
the extra's name so each line still stands alone.

Refs #200

* fix(doctor): show integrity issues as rows under Integrity

Tampered skills and missing hashes printed as free "! Skill integrity:
N skill(s) ..." lines after the verified row, breaking the aligned
Environment rows. They now sit under a single Integrity row, which
turns to ! when any skill has an issue.

Refs #200

* fix(doctor): pluralize the Skillignore and Trash rows

The rows read "1 patterns, 1 skills ignored" and "1 item(s) (50 B),
oldest <1 day". They now read "1 pattern, 1 skill ignored" and "1 item,
50 B · oldest under a day". The JSON messages keep their wording, since
scripts may match on them.

Refs #200

* docs: show the restyled output of batch 3 commands

The reference pages still showed the old output of upgrade, new, trash,
collect, audit, search, backup, restore, extras, hub, target and doctor:
divider headers, arrow-prefixed lines and "(s)" plurals. The examples
now match what the commands print, in all five locales, and the doctor
check table uses the new row labels.

Refs #200

* feat(trash): let trash empty skip its confirmation with --force

Scripts had no way to empty the trash: off a terminal the confirmation
reads stdin and cancels on EOF. --force (-f) skips it, as it does for
uninstall and extras remove.

Refs #200

* fix(doctor): show source, link and target problems as rows

Missing sources, unsupported links, folders without SKILL.md, unknown
skill targets, bad include/exclude patterns and broken symlinks printed
as free "✗ Source not found: …" lines between the aligned rows. They
now use the row of the check they belong to, and broken symlinks or a
missing source add a Next suggestion.

Refs #200

* fix(ui): leave out durations that would read 0.0s

Fast commands ended with "· 0.0s", which says nothing. Every result line
now formats its time through ui.Took, which init already used for its
steps: durations under 0.05s are left out.

Refs #200

* fix(ui): print dev builds without a v prefix

Version labels were built as "v" + version everywhere, so a dev
build printed "vdev". VersionLabel adds the prefix only when the
version starts with a digit, and the logo, help header, version
command and upgrade output share it.

Refs #200

* fix: pluralize category and drop "just now ago"

The checklist status bar appended "s" to its item name and printed
"categorys". The trash TUI formatted ages with its own helper that
added "ago" after "just now". plural now handles consonant + y,
and the trash TUI uses the shared timeAgo, which makes formatAge
unused.

Refs #200

* fix(target): drop the parenthesized counts in the plain list

The plain target list printed "merged (43 shared, 0 local)" while
target info and doctor print "merged · 43 shared" with zero counts
left out. The sync summary now returns both forms from one scan, so
the TUI and --json keep their existing text and the plain list
matches the other commands.

Refs #200

* test(backup): make the agents backup test back something up

The test only synced agents, which become symlinks into the source
and are skipped by backup, so it passed without any backup being
written. It now adds a target-local agent and checks that the file
lands in the backup.

Refs #200

* refactor(cli): render every command's --help with printHelp

About forty help pages were hand-aligned raw strings with "Usage:",
"Options:" and "Examples:" headings, while hooks, mcp, plugin, extras
and the top-level help already used printHelp. They now share it, so
headings, option order (short flag first) and alignment match across
commands. Examples align on their own column so long commands no
longer push option descriptions to the right, and "--help, -h" rows
are dropped as on the converted pages.

Refs #200

* feat(enable): print enable and disable results as rows

enable/disable still used the old "Disabled: x (added to .skillignore)"
line plus a "Run skillshare sync" sentence. They now print one row per
pattern, a Next hint after a change, and the dry-run closing line,
like the other restyled commands.

Refs #200

* fix(sync): pluralize the leftover cleanup note

Project sync printed "Cleaned 1 leftover skill(s)". The note now
uses the singular or plural noun, and the sync docs show it as the
dim note it is instead of an arrow line.

Refs #200

* fix(cli): drop the arrow info lines left in install, update and sync

These commands were restyled, but about fifty messages still went
through ui.Info, which prints a leading arrow, and several said
"skill(s)" or "item(s)". Hints now print as dim notes, empty results
("No agents found") as closing lines, follow-up commands as Next,
and counts use the singular or plural noun.

Refs #200

* fix(cli): tidy small leftovers found in the recorded output

doctor printed "0 minutes ago" for a fresh backup and an ASCII arrow in
the Update row, unlike trash and upgrade, which say "just now" and
"v1.2.0 → v1.3.0". The "bring the Targets up to date" hint capitalized
a word mid-sentence, and search --list showed hub sources under the
home directory as a cut-off absolute path instead of folding it to ~.
JSON messages keep their wording.

Refs #200

* fix(search): fit a home-folded source path in the list

Folding the home directory to ~ was not enough: a typical local hub
path such as ~/.config/skillshare/skills/frontend/react is 42
characters, so the 40-character cap still cut it. Fifty characters
next to the 24-character name column stays within 80 columns.

Refs #200

* feat(list): print the plain list with rows and a closing count

The --no-tui list still used the old header with a divider and an arrow
on every skill, so it looked unlike every other restyled command. Skills
now line up in one column under their group, tracked repos use the
✓ / ! marks, and the list closes with a count split by tracked, remote
and local (zero counts left out). Project mode keeps its scope as a dim
"· project" after the title.

Global and project had copied the same plain-output code; both now call
printSkillList. The loading spinner clears itself instead of leaving a
"Loaded N skills" line.

Refs #200

* fix(cli): show help for subcommand --help instead of running it

target add/remove/list, trash list/empty and hub list/remove passed
--help through as an argument, so they ran the command, failed on a
target or hub named "--help", or rejected the flag. tui on --help even
switched the TUI on. Each now prints its command's help page.

The tui status and toggle lines also move to the row and Done style.

Refs #200

* feat(diff): print differences as rows and close with next steps

The plain diff still printed a Summary line, divider headers, arrow
hints and per-target "scanning" lines when output was not a terminal.
Each group of targets now lists one row per kind of change ("New  a,
b"), targets in sync share a single ✓ row, extras use the same rows,
and diff closes with a count and up to three commands. The suggested
sync command follows what differs: sync, sync agents, sync extras or
sync --all.

Agents no longer show as "New 1 skill", because rows name the change
rather than counting skills. "skillshare diff all", listed in the help,
failed with "target 'all' not found"; it now diffs everything, the
same as no argument.

Refs #200

* test: sync the install warning summary and gofmt two status tests

The "6 findings" wording changed in an earlier commit, but the unit
test under cmd/ still expected "6 finding(s)": scripts/test.sh and
make check only run ./internal and ./tests, so the cmd tests never ran.
Two status tests also had comment alignment gofmt rejects, which fails
make check's fmt-check.

Refs #200

* test: run the cmd unit tests in scripts/test.sh

CI runs go test ./cmd/skillshare/... ./internal/..., but the local
runner behind make test and make check only ran ./internal/..., so a
stale assertion in cmd/skillshare passed locally and would only fail
in CI.

Refs #200

* feat(cli): restyle commit and analyze output

commit and analyze were the last git and inspection commands still
printing the old header and table layouts. They now use the same rows,
Done line, and Next hints as push and pull, so a commit reads like the
commit row of push, and analyze shows each target group as a bold name
with aligned token rows instead of a boxed table.

Refs #200

* feat(log): show entries as rows and fold the log path

The plain log view still opened each section with a boxed header that
printed the absolute log path, then a TIME | CMD | STATUS table. Entries
now read like every other command: a status mark, the command, when it
ran and how long it took, with details indented beneath. The log path is
folded to ~ and moves to a dim footnote.

log --stats drops its divider and ends with one summary line. It and
the TUI stats panel said "just now ago" for a fresh entry; both now use
timeAgo, which already handles that case.

Refs #200

* feat(cli): restyle completion, ui and leftover arrow lines

completion --install, ui start/stop and a few notices still printed
arrow info lines with absolute paths. They now end with a Done line,
fold paths to ~, and give config-file lines their own bold heading so
the lines to paste read apart from the hints. The empty extras TUI now
says the same thing as extras list.

Refs #200

* feat(init): report git changes on an existing setup as rows

init --remote and init --git-root on an existing setup still printed
arrow lines with absolute paths and spinner results in a different
voice. They now print aligned Git, Remote, Pull and Git root rows like
a fresh init, end a dry run with the shared Dry run line, and give the
choices after a remote check as Next commands.

A source repo the user created before init had no fallback identity,
so on a machine without git config the follow-up commit failed and
dumped git's whole "Please tell me who you are" message. The existing
repo now gets the same local fallback a new one does.

Refs #200

* feat(doctor): warn when a newer built-in skill is published

status already compared the skillshare skill with the published one,
but doctor only printed the local version, so a run meant to find
problems called an old skill healthy. doctor now fetches the published
version alongside the CLI update check and suggests upgrade --skill.

Both commands share a semver comparison. status compared with !=, so a
skill ahead of main (a dev build) was reported as having an update.

Refs #200

* fix(cli): count batch audit warnings in words and clear progress spinners

The batch install and update audit summary still printed "2 skill(s)"
and "warning(s)" after the rest of the output switched to "1 skill" /
"2 skills". uninstall and agent update also left "Found N skills" and
"Check complete" behind, progress messages the restyle clears.

Refs #200

* fix(cli): say so when diff or analyze has no targets

With no targets configured, diff reported "No differences" and analyze
printed only its "Add -v" hint, both reading like a healthy result.
They now say no targets are configured and suggest target add, as
target list does.

Refs #200

* fix(diff): align values across target sections

Each target section sized its labels on its own, so a section with
"Local only" started its values two columns later than one with only
"New", and the list read ragged. All sections now share one width.

Refs #200

* fix(doctor): count built-in skill warnings in the summary

The skill rows (not found, missing version, newer version available)
were marked as warnings and counted by --json, but never reached the
text summary, so doctor could print "! Skill not found" and close with
"All checks passed". The all-good integration test only passed because
of this; it now installs the built-in skill it needs.

Refs #200

* fix(init): stop writing a git identity into existing repos

On a repo the user created, init stored skillshare@local in the repo
config so its one commit would work. That setting outranks --global,
so the advice we printed never took effect and later pushes went out
as skillshare@local. The commit now gets the fallback through env
vars only. The note for repos init creates names the repo-level
command that actually overrides it.

Dry runs also wrote .gitignore into existing repos while closing with
"nothing was written", and warned about --no-git when it was not
passed; both now only describe what would happen.

Refs #200

* fix(cli): tell "no targets" apart from "nothing reaches them"

diff and analyze said "No targets configured" and suggested adding one
whenever there was nothing to show, including when targets exist but
have skills turned off or no skill is meant for them.

Refs #200

* fix(log): drop the trailing separator on instant entries

Entries with no recorded duration ended in a dangling " · ". Also
removes the formatRelativeTime alias the stats restyle left unused;
its test now covers formatDurationShort directly.

Refs #200

* docs: match tui, project list and log examples to the real output

The tui and project list examples used label widths the CLI does not
print, and the plain log example claimed to show non-TTY output, which
joins each entry's details on one line.

Refs #200

* perf(version): cache the published skill version for a day

doctor and status asked GitHub for the latest skill version on every
run, adding up to 3 s when offline. They now reuse the answer for 24
hours, the same interval the CLI update check caches for. upgrade
--skill still fetches fresh and records what it got.

Refs #200

* perf(server): reuse the cached skill version on the dashboard

The version endpoint fetched the published skill version on every
call, like doctor and status did, so it now shares their 24-hour
cache. It also treated any difference as an update, offering a
"newer" skill to anyone running a local build ahead of main.

Refs #200

* test(redteam): match the restyled update --diff heading

update --diff now titles its file list "Files changed" in sentence case,
like every other section, so the check for the list looked for the old
"Files Changed" box title and failed although the list was there.

Refs #200

* test(cli): keep /tmp fixtures unfolded where HOME is /tmp

scripts/test.sh now runs the cmd unit tests, so the docker sandbox runs
them too, with HOME=/tmp. shortenPath then folds the /tmp fixtures in
four tests to ~ and their exact-path checks fail. The tests now point
the cached home elsewhere while they run.

Refs #200

* fix(init): find skills/ in a repo that holds nothing else

The check for a skills/ folder ran only when the repo root had no skill
folders, but it counted skills/ itself as one, so it never ran. A repo
with only skills/ became the source as a whole, one level too high.

Refs #200

* fix(init): keep the skills git scope for a custom source

With a remote that has no skills yet, init defaulted to versioning the
whole skillshare folder. For a --source outside that folder it holds
none of the user's skills, so the commit and every later push carried
nothing while init reported the remote as linked.

Refs #200

* test(sync): check only the Agents section for updated agents

The restyle turned the old "0 updated" check on the agent summary into
"no 'updated' anywhere", which also covered the Skills section. On
macOS the first project sync reports the skill as updated there, so
the test failed on something it was never meant to check.

Refs #200
2026-10-03 21:43:35 +08:00
Willie 5d269e8e4c feat(push): add --pull for two-way sync in one command (#354)
* feat(push): add --pull for two-way sync in one command

Users editing skills on several machines have to remember push on one and pull on the other, and a push from a machine that is behind fails until they pull first. push --pull commits local changes, merges the remote with the same merge semantics pull uses, pushes, and syncs the git root scope.

A conflict stops before anything is pushed and keeps the local commit; a target sync failure after a successful push is reported separately so the user knows the remote is already updated. It never rebases or force-pushes. commit rejects --pull instead of silently ignoring it, since both commands share the argument parser.

* docs(push): document push --pull across guides and translations

Cover the new flag on the push reference, the cross-machine sync guide, the quick reference, the daily workflow and the cross-machine recipe, in English and the ja, ko, zh-Hans and zh-Hant translations, plus the built-in skill's sync reference.

* fix(push): print the scope's sync command when --pull sync fails

A bare 'skillshare sync' retry hint defaults to skills and may pick project mode, so it does not redo an agents, extras, or root sync. Derive both the sync and the hint from one per-scope list.

* fix(pull): sync extras for the root git scope

The root scope versions skills, agents and extras, but pull and push --pull only synced skills and agents, so merged extras changes never reached their targets.

* fix(server): sync extras after a root-scope pull

The root scope versions extras too, but the dashboard pull only synced skills and agents, so merged extras never reached their targets. Matches the CLI fix.

* feat(ui): add a Sync both ways action to Git Sync

Mirrors skillshare push --pull: commit local changes, merge the remote and sync targets, then push, in one click. A pull conflict stops at the existing review before anything is pushed.

* feat(ui): confirm Sync both ways and list what it will do

Sync both ways commits, merges and pushes in one click, so ask first and show the concrete outcome (files to commit, remote commits to merge, scope, push target, conflict behavior) before changing anything.

* fix(ui): make Sync both ways match push --pull

It skipped target sync when the pull was already up to date, stopped on an empty remote instead of pushing, and the pull dropped extras warnings. Add alwaysSync and a remote_empty code to /api/pull and surface all extras warnings.
2026-10-03 18:53:08 +08:00
Willie 6e176ddb0c feat(extras): share Markdown memory across agents (#349)
* feat(extras): add shared Markdown memory management

Store user-managed notes in extras/memory for global and project scopes.
Add CLI operations, a nested dashboard browser, versioned edits, and
backup-backed deletion and recovery.

Connect selected agents with reviewed, marked reading guidance while
preserving existing instruction files and routes. Add explicit INDEX
linking, conflict draft comparison, and copy-content previews.

Document the workflow in five website locales with English screenshots
and translate dashboard copy across all eleven locales. Native automatic
memory remains separate from this shared file overlay.

Validation: make check; 22 UI tests; TypeScript and ESLint; production UI
build; six isolated CLI E2E steps; five-locale website build; explicit
Claude-write/fresh-Codex-read smoke test.

* feat(memory): support dashboard note moves and redesign browser

Add version-checked rename and folder moves without replacing existing notes.
Back up the old path and preserve content and permissions while refreshing
selection, search, index warnings, and backup links.

Integrate the two-pane Memory browser and compact agent controls. Refresh all
English screenshots and document move/link/history behavior in five locales.

Validation: make check, scoped UI tests, TypeScript, ESLint, UI production build,
five-locale website build, locale parity, and visible Chrome workflow checks.

* fix(memory): recheck reviewed files before committing changes

Preserve external edits detected after note backup or draft preparation, and create new notes exclusively. Revalidate each guidance file before writing, report stale partial failures, and correct the built-in move/rename documentation.

* fix(memory): repair initialization and preserve note recovery scope

Detect missing starter files before dashboard note creation and refresh the index after initialization. Honor requested backup paths, include external project memory sources in recovery scope, and log initialization failures after partial mutations.

* fix(memory): protect guidance creates and match Windows history paths

Create new reviewed instruction files exclusively to preserve competing writes. Match Windows history links with native backup paths while keeping POSIX names and backup API paths intact.

* fix(memory): require configured extra for external backup access

Keep project backup scope inside the repository unless a valid memory folder extra enables its external source. Cover listing, version preview and restore denial without changing configured-source recovery.

* fix(memory): lock backup scope validation through restore

Validate project memory backup access under the restore write lock so concurrent extra deletion cannot revoke access between validation and writing. Hold read locks across scoped backup reads as well.
2026-10-03 14:16:58 +08:00
Willie 1a1eeee844 fix(check): compare project-relative local sources and record install layout (#337)
* fix(check): compare project-relative local sources and record install layout

Project installs record a local source as typed (./vendor/skill), relative
to the project root, so check skipped them. Resolve them against the project
root in both the CLI and the dashboard.

The copied shape of a local install was inferred from its file hashes, so a
CLI install whose source held only SKILL.md looked like a dashboard
collection-root install: a child skill added at the source later was neither
reported by check nor copied by update. Record the layout (skill-file or
directory) at install time and keep the hash heuristic only for entries
written before the field existed.

* fix(install): let a recorded layout decide the update shape without hashes

An entry whose hashing failed at install time still knows what was copied;
returning false there would turn a SKILL.md-only update into a full copy.
2026-10-02 20:56:55 +08:00