Commit Graph
820 Commits
Author SHA1 Message Date
Willie 10466f2d53 fix(server): report one target per tied sync budget warning
When several targets tied for the worst context cost, the warning joined
their names but listed top offenders from whichever target was seen
first, so the offenders could belong to a different target than the
one a reader looks at. Map iteration order also made that choice
random.

Ties now resolve to the alphabetically first target, and the warning
carries that target's own offenders. The test gives each tied target a
different skill so a mismatch between target and offenders shows up.
2026-09-30 00:24:51 +08:00
Willie 8962ed9416 fix(server): name the target in sync budget warnings
Context-cost warnings in the Web API's sync response never carried a
target name: the warning struct tracked the worst token totals across
targets but dropped which target hit them. The CLI's sync --json and
the dashboard's ContextCostWarning both include target, so the
dashboard's i18n string rendered its {target} placeholder verbatim
instead of the target's name.

computeContextCost now records the names behind the worst always-loaded
and on-demand totals, joining ties with a comma, and reports them as
the warning's target. Regression tests cover both the tied case and a
worst-offender that only one target receives.
2026-09-29 23:30:21 +08:00
Willie fd3bc824f3 feat(targets): edit other files a target's tool reads from its page
Pi and OMP read APPEND_SYSTEM.md next to their skills, and editing it
meant leaving the dashboard. A target page now shows extra file tabs:
built-in ones come from targets.yaml `files:` (APPEND_SYSTEM.md for pi
and omp), and users can add their own, including files in subfolders.
Paths stay inside the tool's folder; three tabs show and the rest move
to a menu. "Share with Extras" opens Add extra prefilled with the file's
folder and name, so the same file can be shared across tools.

A single-file extra's name now defaults to its file name without the
extension (APPEND_SYSTEM.md -> APPEND_SYSTEM) until the user types one,
and the hint suggests the file name, as the #300 reporter asked.

Refs #301
Refs #300
2026-09-29 15:25:06 +08:00
Willie 0aa6900050 fix(mcp): take over an Agent entry that already matches on sync
Importing a server from an Agent without ticking that Agent, then ticking
it later, left the Agent's entry unmanaged for good: the plan reported it
as unchanged, recorded no ownership, and so unticking the Agent had
nothing to remove. Nothing on the page said why, because the server's
name was now defined in the source.

Such an entry is now planned as adopt. Sync records it as managed without
rewriting the file, so the takeover is visible in the preview and happens
only when the user syncs. A server the person turned off in the Agent
itself is still never claimed.

Refs #303
2026-09-29 15:23:07 +08:00
Willie 0931078e2a fix(targets): stop treating Antigravity as a reader of ~/.agents/skills
Antigravity's docs list only ~/.gemini/config/skills as the desktop app's
global skills folder, and an independent test found it does not read
~/.agents/skills. The entry made the dashboard claim Antigravity still
sees universal's skills after switching its own skills off.
2026-09-29 11:45:48 +08:00
Willie e48784f16b feat(sync): name targets that sync one folder with different filters
When two targets sync skills into the same folder with different include
or exclude filters, each sync adds what one wants and prunes what the
other filters out, so the folder never settles and sync keeps showing the
same changes. The only hint was a vague overlap count.

sync, doctor and the dashboard's Sync page now name the targets and the
folder, and suggest turning skills off for all but one (universal when it
is among them) instead of removing a target. The Sync page offers the
existing stop-syncing dialog in place. Targets with identical filters are
not reported.
2026-09-29 11:45:48 +08:00
Willie 34db034262 fix(targets): stop reporting a target's own skills folder as another source
A codex target pointed at ~/.agents/skills shares that folder with
universal, so it holds one copy of each skill. The target page still said
codex also reads universal's skills and every skill shows up twice. A
folder the target itself syncs into no longer counts as another source;
with skills off, reading universal's folder is still reported.
2026-09-29 11:45:48 +08:00
Willie deafea7693 fix(plugin): allow Codex logos up to 1MB
Real plugin logos exceed the 256KB cap (ponytail ships a 554KB PNG), so
the dashboard fell back to the default icon for them.
2026-09-29 11:29:41 +08:00
Willie a996a5bbfd feat(plugin): show version and Codex logo for plugins with no Agent
A plugin added without choosing an Agent has no binding, so the
dashboard had nowhere to read its version from and always drew the
default icon.

Record the source version on the package when it is added, and let
rows without one ask their source once. Read interface.logo from the
Codex manifest (inside the plugin, image types, 256KB max) and draw
it in the list and in the add preview, falling back to the default
icon.
2026-09-29 11:25:50 +08:00
Willie 01f7bc65cd fix(targets): give kilocode its documented AGENTS.md location
Kilo Code had no instruction file in the target list, so the dashboard
guessed ~/.kilocode/AGENTS.md from the target name, a folder Kilo never
reads. Kilo documents ~/.config/kilo/AGENTS.md globally and AGENTS.md at
the project root, which both the CLI and the VS Code extension read; a
custom path stays available through the target's instructions setting.

Refs #302
2026-09-29 10:01:19 +08:00
Willie d757afbb61 feat(extras): let several single-file extras share one source folder
A single-file extra added from the dashboard with the shared source got
<extras dir>/<name> as its folder, so a second file in the same folder
needed the same name and was refused as a duplicate. Project mode had no
source setting at all, so it could never share a folder.

The add dialog now has a source folder field that defaults to the name,
and project extras accept a source relative to the project root, from the
dashboard and from `extras init -p --source`. Paths shown for single-file
extras also keep the folder's own separator, so Windows paths no longer
mix backslashes with a slash before the file name.

Refs #300
2026-09-29 09:43:29 +08:00
Willie e0882217ab feat(targets): list copy-mode copies apart when skills are switched off
Switching skills off keeps copy-mode copies, since they are real folders the
user may have edited, but they were listed as the user's own items. The tool
still loads them next to wherever it now reads skills from, so each skill
shows up twice. The CLI, the preview and the dashboard dialog now list them
separately and say to delete them if the tool reads the same skills
elsewhere, leaving that choice to the user.
2026-09-29 05:13:06 +08:00
Willie 7bc8bd6634 fix(targets): count what a skills-off folder still holds and refuse mixed flags
With skills off, analyze only counted unmanaged folders, yet the tool still
loads the links a target sharing the folder keeps and the copies copy mode
leaves behind, so the report showed fewer skills than the tool reads.

`target <name> --skills=false` combined with include/exclude flags applied
the filters and silently dropped the switch, because filter updates return
first. The switch removes links, so the mix is now refused instead.
2026-09-29 05:03:49 +08:00
Willie 46b23cd2ad feat(targets): let a target keep its instructions with skills switched off
Some tools only need skillshare for AGENTS.md, MCP or agents, while their
skills folder is managed elsewhere or shared through another target.
Syncing skills into those folders duplicated what the tool already reads,
so a target can now turn skills off (`target <name> --no-skills`, the
dashboard switch) and sync, diff, status and doctor skip its skills.
Switching off warns about the other tools that read the same folder.

The target list also gains tools that were missing, with colored vendor
logos where their licenses allow it, and zencoder/replit default paths
follow their current documentation.

The dashboard's target pages get the matching UI: universal is pinned as
the shared folder when adding a target, the add dialog lists what will be
written, and the AGENTS.md tab shows where the file comes from in one card
with an in-place switch between shared files.
2026-09-29 04:51:24 +08:00
Willie 00f51adbda feat(extras): create and manage single-file extras from the CLI and dashboard
Single-file extras (file + per-target as) already worked in the sync
engine, but the only way to create one was editing config.yaml or the
dashboard's AGENTS.md tab, so they looked like an AGENTS.md-only feature.
Syncing a file such as Pi's APPEND_SYSTEM.md required hand-written config.

- extras init gains --file and --as; the init TUI asks Folder or Single
  file. Validation runs before anything is written, and init neither
  creates the source file nor syncs.
- extras list and add-target show the full source and target file paths.
- POST /api/extras accepts file and per-target as, with the same
  connection checks as add-target.
- The dashboard's first Extras tab becomes "Folders & files" and lists
  every extra except AGENTS.md files; Add extra can create a single file
  with a per-target file name.
- Assigning shared AGENTS.md files now only attaches or detaches AGENTS.md
  extras. Before, saving on the AGENTS.md tab restored and removed any
  other single-file extra linked to the same instruction file.

Refs #300
2026-09-29 00:27:48 +08:00
Willie 13a93242f8 feat(ui): redesign the Hubs page around one list with in-place editing
Previewing a hub from the builder landed on an empty "pick a hub" state:
the builder and the browser kept separate copies of the drafts and the
browser's copy went stale. Merging both into one page removes the split,
so your own hub is shown exactly as others will see it and edited in place.

Hub entries can now be added from any repository URL, not only from
installed skills, with a branch or tag picked from the remote. The version
is written into the source URL, the only place the hub format keeps it,
and a new /api/hub/refs endpoint lists refs and rewrites sources so the UI
never builds tree/<ref> URLs itself. Skills added from installed ones keep
the version they were installed from.

Draft reads and saves read refs from the URLs only, so opening a hub never
waits on a remote. Added hubs get unique labels, since the default hub is
stored by label.
2026-09-28 23:39:19 +08:00
Willie 529a8f2771 fix(install): keep config installs on the recorded branch
Installing from config set the branch only when one was recorded, so a
legacy entry with a tree/<tag>/... source and no branch was pinned to
that tag, while update kept it on the default branch. Use
ApplyRecordedBranch like the update paths do. Settling the ref up front
also lets web URL entries from one repo share a single clone again.

Refs #293
2026-09-28 22:12:09 +08:00
Willie 72b6a38dfd fix(install): treat tree/HEAD web URLs as the default branch
GitHub links to tree/HEAD/... when it means the default branch. Since
URL refs became clone refs, HEAD was taken as a branch name: with a
subdir the ref lookup failed, without one git clone --branch HEAD
failed. These URLs installed fine before, so keep them on the default
branch.

Refs #293
2026-09-28 22:12:09 +08:00
a352b8d243 feat(install): pin skills to a tag or commit from their web URL (#293)
* feat(install): pin GitHub tree/blob URLs to their ref

Hub entries only carry a source URL, so a team hub had no way to pin a
third-party skill to a reviewed tag or commit. The parser already read
tree/<ref>/ and blob/<ref>/ URLs but discarded the ref, installing from
the default branch instead of the version the URL names.

Keep that ref as Source.Branch. An explicit --branch (CLI) or branch
(API body) still overrides it; the server handlers now only override
when a branch is given. Batch installs group by repo and ref so two
pinned versions of one repo get separate clones.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(install): describe pinning hub sources to a tag or commit

Explain that a GitHub tree/<ref>/ source URL pins the install, in both
the hub index guide and the install reference.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* feat(install): pin GitLab/Bitbucket URLs and resolve refs containing "/"

Pinning should not depend on the host, so GitLab (-/tree/<ref>/) and
Bitbucket (src/<ref>/) web URLs now keep their ref too.

Taking the first path segment as the ref broke branch names containing
"/": tree/feature/x/skills/foo cloned branch "feature" and failed. When
the first segment is not a branch or tag on the remote, the longest
matching one is used instead. A ref the remote does not have fails
with a hint to use --branch, rather than silently installing the
default branch. A saved branch that covers more of the path (from
--branch or config) moves the subdir without asking the remote.

The grouped-install round-trip test pinned -/tree/main/ against a repo
on master; the ref used to be ignored, so its branch is now renamed.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* docs(install): cover GitLab, Bitbucket and "/" branches in URL pinning

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(install): keep web URL refs correct on reinstall, --branch and grouping

Review of the slash-branch resolution found four gaps:

- A saved branch equal to the whole URL tail (tree/feature/x) did not
  move the subdir back, so every reinstall or update looked for "x".
- An unrelated --branch with a slash-branch URL kept the parse-time
  subdir (x/skills/foo) instead of the URL's real one.
- Grouped installs copy the source to a repo root; resolution then put
  the subdir back and discovery looked under a doubled path. Grouping
  also keyed on the unresolved first segment, so tree/feature/x and
  tree/feature/y could share one clone. Web URLs with a ref now install
  on their own, and resolution leaves a repo-root copy at its root.
- A skill named after the parse-time subdir kept that name once the
  path turned out to be all ref (tree/feature/x named the skill "x").

The unknown-ref error now reads as a missing ref first, since a stale
ref (tree/master after a rename) is more common than a slash branch.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(update): keep legacy installs on the default branch

Skills installed before URL refs were honoured got the default branch
and recorded no branch. Single-skill update re-parses the recorded
source, so it moved them to the URL's ref (or failed on a stale one,
with a hint to pass --branch, which update does not accept), while
update --all rebuilds from RepoURL and kept the default branch.

ApplyRecordedBranch treats an empty recorded branch as the remote
default and drops the URL's ref. It replaces the branch handling at
every site that reinstalls from recorded metadata: CLI and server
update for skills and agents, and tracked-repo rehydration.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* perf(search): share one clone across pinned hub skills from a repo

Any web URL with a path after its ref installed on its own, so a hub
listing many tree/<tag>/ skills from one repo cloned it once per skill.

groupByRepo now resolves each URL's ref before grouping, with a
RemoteRefs cache that lists each remote's refs once, and groups by
repo and resolved ref. Only refs containing "/" (whose subdir the
discovery match cannot re-derive) and refs that fail to resolve still
install alone. A resolved source is marked settled so the grouped
clone does not resolve it again, and ApplyRecordedBranch settles the
subdir from a recorded branch without asking the remote.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Willie <egg358358@gmail.com>
2026-09-28 22:08:16 +08:00
Willie d572ba3b32 perf(backup): walk each snapshot once when listing backups
The list endpoint walked every snapshot to size it, then walked all of
them again for the total. With thousands of files and a cold file cache
(right after an upgrade, for example) the Backup page took visibly long
to open.

Each snapshot folder is now walked once for both its size and file
count. Backup and total sizes are sums of those, and the per-folder
counts are returned as `entries` so the dashboard can show what each
folder holds.
2026-09-28 21:51:11 +08:00
Willie 86dca7924f feat(backup): browse and restore file history, delete snapshots
skillshare backs up AGENTS.md, CLAUDE.md and shared-file locations before
rewriting them, but nothing could list or restore those copies, and a
target snapshot could not be deleted on its own. The file backups now
record why they were taken in their name (<time>.<reason>.bak; older ones
have none), and the server and CLI can list them, show a version and
restore it, saving the current content first. A link is only replaced
when the caller asks for it.

Also:
- DELETE /api/backups/{timestamp} and `backup --delete`.
- Agents snapshots (<target>-agents) could not be restored from the
  dashboard: the name was looked up as a target.
- Project mode uses the project's backups folder for snapshots and lists
  only file backups inside the project.
- MCP backups report which servers they changed and when.

Refs #299
2026-09-28 19:23:16 +08:00
Willie 342fac7691 feat(server): serve shared files and their locations in project mode
In a project, single-file extras were only reachable through the folder
extras API, which has no file name, no import mode and no immediate
sync. The dashboard now shows them on the AGENTS.md tab like global
shared files, so the instructions routes it needs must work in project
mode: list, create, content, resolve and the locations routes.

A project location must stay inside the project (absolute, ~ and ..
paths are refused with instructions_location_outside_project) and is
stored relative, so a clone keeps working. A tool's own file is a valid
location in a project, so the is_target check is skipped there.

Refs #299
2026-09-28 18:40:15 +08:00
Willie 0bbcbd7d54 fix(server): settle and roll back instruction locations
A location whose file was edited outside skillshare had no way to collect
or reapply, and a location whose config save failed stayed written on
disk. Resolve now takes a location path, and a failed save restores
what the file held; a failed reload keeps the saved location.

Refs #299
2026-09-28 17:41:33 +08:00
Willie 04485099ed feat(server): add locations API for shared instruction files
The dashboard could only attach a shared AGENTS.md to known tools, while
the CLI's --add-target/--as already writes it to any folder under any
name. Add list/add/mode/remove/restore-preview routes for such locations,
syncing a new location right away and refusing folders in the way, tool
files and files held by another shared file. The extras add-target route
now accepts as and runs the same config validation as the CLI.

Refs #299
2026-09-28 17:30:41 +08:00
Willie 7baf716c63 feat(instructions): translate dashboard warnings and errors
Mode changes, attach and the other AGENTS.md actions returned English
sentences, so a zh-TW dashboard showed "backed up ... before replacing
it" and "... is a directory" untranslated. Warnings are now
{code, params, message} and every instructions error carries a code,
so the dashboard shows them from i18n and falls back to the English
message for unknown codes. CLI output is unchanged.

Also from the Windows re-test:
- replacing a user's junction says so and where it pointed;
- status, doctor and sync agree on identical unowned copies (kept, not
  linked), and an identical copy no longer suggests --force;
- the targets list shows a Developer Mode tooltip when file links are
  unavailable;
- file names are taken from Windows paths too.

Refs #299
2026-09-28 16:12:59 +08:00
Willie 0ae1e17e15 fix(instructions): say when a folder blocks a target, pluralize kept files
A real directory at a target path showed the generic drift hint ("sync
will change it back") although sync refuses to replace a directory. The
assignment now carries reason "directory" and the row says to remove or
rename the folder.

The restore plan read "personal stay" in English because the kept-files
line had no singular form.

Refs #299
2026-09-28 15:41:35 +08:00
Willie 87e6cc9538 fix(extras): guard shared-file ownership and keep user files on restore
Exploratory CLI, dashboard and real-Windows testing of single-file extras
found paths that corrupted another shared file, lost user content, or
reported the wrong state.

- One shared file per link/copy target is now enforced in config
  validation, so the CLI, sync and API all refuse it; an import is never
  written through a link into another shared file (409
  instructions_target_held in the API).
- Restore recreates an original Windows junction as a junction, which
  needs no privilege, and remove restores before saving config so a
  failed restore no longer orphans the record.
- Imports into dotfiles links, damaged managed blocks, repointed links
  and CRLF files keep the user's content; project imports use paths
  relative to the file so a moved repo keeps working.
- Mode and assign responses carry sync warnings; dry-run says when an
  edit would be backed up; invalid modes no longer show as healthy.
- An edited copy reports modified, identical unowned copies count as
  synced without being owned, and relative project links no longer show
  drift.
- sync extras --json exits non-zero on errors; extras <name> --help
  works; --add-target gains --as; detach messages match behavior.

Refs #299
2026-09-28 15:27:02 +08:00
Willie f307c8a114 fix(sync): keep user files safe in copy fallback and mode switches
Review of the file-link fallback found paths that lost or misreported
user data:

- An identical local file was adopted as a skillshare copy and later
  pruned without backup. Only files sync writes are tracked now.
- Clearing all own text in import mode was treated as no record, so
  switching back resurrected the attach-time content.
- The mode API saved the new mode and returned success when a directory
  blocked the sync; it now keeps the old mode and returns 409.
- Restore preview promised a delete for files restore leaves alone; both
  now share one ownership decision.
- Prune forgot ownership even when removing the copy failed.

Refs #299
2026-09-28 14:27:53 +08:00
Willie c6ec545e96 feat(instructions): update copies when a shared file is saved
Saving a shared AGENTS.md in the dashboard only wrote the source, so every
target in copy mode (including the Windows copy fallback) kept the old
content until someone found the sync button. Saving now rewrites that
file's copy targets through the regular extras sync, so an edited copy is
still backed up first, and the toast names the copies it updated. Link and
import targets already read the source and are left alone.

Refs #299
2026-09-28 14:08:46 +08:00
Willie faa5186818 fix(sync): keep import-mode edits when switching back to import
Switching a target from import to copy or symlink and back rebuilt the
file from the attach-time restore point, so lines the user wrote while in
import mode vanished (they survived only as a drift backup). Save the
file's own content when leaving import and rebuild from that on the way
back; the attach-time restore point itself is never re-recorded.

Refs #299
2026-09-28 14:04:15 +08:00
Willie 3f4d347e61 fix(ui): honor NO_COLOR in every command's output
The legacy color constants in internal/ui (Red, Gray, Dim, Bold, Reset, …)
were compile-time strings that ignored the theme, so status, doctor, extras
list and about twenty other commands printed ANSI codes even with NO_COLOR
set. Make them variables and blank them at startup when the theme reports
NoColor, so the theme stays the single source of that decision. The two
places that hardcoded escape codes (usage and log --stats) now use the same
variables.
2026-09-28 13:56:59 +08:00
Willie 88f5201911 feat(instructions): choose sync modes, preview restores, and move files
The AGENTS.md tab picked each target's mode for the user (import where
supported, a link elsewhere), which left Windows users without Developer
Mode no way out of a broken link and gave no copy option. Restore replaced
a file without showing what would come back, and only custom tools could
change which file a target reads.

- Each connected target gets a mode picker (import, symlink, copy); symlink
  is unavailable when the machine cannot create file links, and a target
  whose file is a folder link says so.
- Restore opens a preview of the file after restore and its diff.
- Every supported target can change its file location and name in a
  dialog, with reset to default.
- The target file tab is a single column with the read order on one line,
  a tinted managed import block, soft wrap, and an edit/preview switch;
  universal lists the files of the tools that read its skills.
- The convert dialog shows only the methods that apply and a ready preview.

Refs #299
2026-09-28 13:55:51 +08:00
Willie c7eb59a78f fix(sync): stop linking single files with junctions on Windows
createLink always tried `mklink /J` first. mklink does not check the
source type, so agents, single-file extras such as a shared AGENTS.md, and
per-file extras links became directory junctions that point at a file:
Explorer shows a folder and tools cannot read it. Sync then compared only
the link destination and kept the broken junction forever.

Files now get real symlinks only. When the OS cannot create them (Windows
without Developer Mode, non-elevated), file-based link modes fall back to
copies at the mode level, so status, doctor, the API and later syncs all
see an effective `copy` mode. Fallback copies are tracked by hash so they
keep updating and pruning without touching the user's own files, and turn
back into links once links work. Single-file extras remember what they last
wrote, so an unedited copy is replaced silently instead of producing a
drift backup on every source edit, and switching a target back to import
restores the user's attach-time content plus the import block.

Includes a Windows runbook and script, verified on a Windows 11 ARM64 VM
with and without symlink rights.

Refs #299
2026-09-28 13:54:52 +08:00
Willie 97499ebcc4 fix: recognise Windows junctions wherever links are detected
Since Go 1.23, os.Lstat reports NTFS junctions as irregular files instead
of symlinks (winsymlink, active for go 1.25 in go.mod). Every check that
looked only at os.ModeSymlink therefore missed the junctions skillshare
itself creates on Windows: linked skills could be treated as local folders,
backed up, copied through, or left behind on prune.

Detect links through utils.IsLinkMode, which also asks the OS about the
reparse tag of irregular entries, and count only real symlinks and
junctions as links so OneDrive placeholders and deduplicated files stay
regular files. restore.go also used os.Stat for its link check, which
follows the link and never matched.

Refs #299
2026-09-28 13:53:16 +08:00
Willie daf77b037f fix(target): keep skills another target still syncs when removing one
Codex and universal both write to ~/.agents/skills. Removing one of them
unlinked that folder, turning the other's synced skills into local
copies it no longer managed. Removal now leaves the folder alone while
another remaining target uses it, in the dashboard, the CLI, the TUI and
the dry-run preview.
2026-09-28 01:58:29 +08:00
Willie ad3d84b76a feat(instructions): manage the files of tools that read universal's skills
With only universal configured, tools such as Codex, Gemini CLI and Pi
still read skills from ~/.agents/skills, but none of them reads
~/.agents/AGENTS.md, and their own instruction files had no page and
could not take a shared AGENTS.md. Universal's AGENTS.md tab now offers
the installed ones in a dropdown, and the shared AGENTS.md list includes
them. A tool is listed once per file and disappears when it becomes a
target of its own.

Cline and the Warp Agent CLI read ~/.agents/AGENTS.md itself, per their
docs, so they now share universal's file.
2026-09-28 01:36:42 +08:00
Willie 3d08cbc670 fix(plugins): name a git failure's cause only when git names it
Git opens every HTTPS failure with "unable to access", a bad certificate
or a 403 included, and a missing git-lfs says "command not found". Those
were reported as a network problem or a missing source and sent the user
to the wrong fix; they now fall back to the general failure.

Discovery of a managed plugin is also cached per plugin, since two
plugins of one source can have snapshots at different commits.
2026-09-27 23:08:39 +08:00
Willie c25eb4f708 fix(plugins): open managed plugins from their snapshot and explain slow reads
Opening a plugin cloned its whole source again to learn which Agents
can take it, which took tens of seconds per row. A managed plugin
already has a reviewed snapshot; it answers when its digest still
matches, and the network is used only otherwise.

A failed clone said only that git failed. Git's output is sorted into
an unreachable source or a missing one without passing it on, since it
can hold credentials. The dashboard translates those, notes when a read
or an Agent's CLI is slow, and says when the browser is offline.
2026-09-27 23:03:21 +08:00
Willie 31697d9bdb feat(targets): know the instruction files of more targets
Targets without an instructions entry asked the user to type the path,
even when the tool documents a fixed one: universal always reads
~/.agents/AGENTS.md and Pi reads ~/.pi/agent/AGENTS.md. Paths for 20
more targets come from each tool's official documentation; targets
whose file is undocumented or changing are left for the user to set.
2026-09-27 23:03:21 +08:00
Willie ad34dbde2d fix(server): load the dashboard when opened at a nested route
The build links its files as ./assets/..., so a page reloaded at
/targets/claude asked for /targets/assets/... and got index.html back.
The browser refused the chunks as text/html and the page stayed on its
skeleton. The server now injects a <base> for the base path.
2026-09-27 23:03:21 +08:00
Willie 5d1f1dc724 fix(mcp): write pi-mcp-adapter servers to mcp-adapter.json
pi-mcp-adapter 3.0 no longer reads Pi's mcp.json, which it leaves to
Pi's upcoming built-in MCP support, and reads mcp-adapter.json in the
same folder instead. Servers synced for the adapter were landing in a
file it ignores, so they stopped loading.

Adapter servers now go to mcp-adapter.json in the global, project and
account folders; pi-mcp-extension still reads mcp.json. The next sync
removes the entries Skillshare had written to mcp.json and leaves the
user's own. If the user already renamed the file as Pi's warning
suggests, Skillshare keeps owning the entries that arrived unchanged
instead of reporting them as someone else's. Import and the unmanaged
scan read both files.

Refs #298
2026-09-27 18:05:26 +08:00
Willie 86fb63fd4b docs: document sharing AGENTS.md from the dashboard
The shared AGENTS.md workflow is dashboard-only, so users had no written
reference for how shared files attach, what restore puts back, or where
drift backups go. Adds a how-to page in every locale, documents
single-file extras and the target `instructions` key, and brings the
translated extras collect section up to date with English.
2026-09-27 04:17:41 +08:00
Willie b8291895e1 feat(ui): share one AGENTS.md across tools from the dashboard
Claude Code now reads AGENTS.md, but every tool still keeps its standing
instructions in its own file (CLAUDE.md, GEMINI.md, AGENTS.md), and Claude
has no user-level AGENTS.md at all. Keeping several tools on the same
instructions meant copying text between those files by hand.

Each target page gets a tab named after the file it reads, with its read
order, an editor, and a convert flow from CLAUDE.md to AGENTS.md. In global
mode the Extras page gets an AGENTS.md tab where several shared files can
coexist and each target picks one or keeps its own. Shared files are
single-file extras: tools that follow @imports get a managed import block
and keep their own lines, others get a link in place of their file.
Projects get small shims instead, since every tool there reads ./AGENTS.md.
Custom targets can declare which file they read.

Restore records the target's state when it is first attached (file,
symlink, or no file) and always returns to it. Edits made after attaching
are kept as drift backups instead of becoming the restore point, so sync,
reapply and mode switches can no longer lose the user's original file.
2026-09-27 03:02:51 +08:00
Willie 9dab28fc19 fix(git): undo a pull's partial checkout and explain permission errors
When git cannot finish a pull's checkout (for example, it may not unlink
a root-owned file), it stops without undoing the files it already wrote,
and with no MERGE_HEAD there is nothing for merge --abort to undo. The
remote's copies then showed up as local changes: the dashboard blocked
the next pull and suggested committing them, which would push the stale
remote content back as a new commit.

A failed pull now restores paths that were clean before it started and
now hold exactly the upstream version, leaving the user's own edits
alone. The dashboard reports permission failures with a
permission_denied code and shows the chown command that gives the
source folder back to the current user.
2026-09-27 02:10:12 +08:00
Willie 779ec26ff5 fix(upgrade): elevate only the binary replacement with sudo
When the binary lived in a root-owned directory, upgrade re-ran the whole
command under sudo. The built-in skill download, UI asset cache, and oplog
then landed in the user's home as root-owned files, and a later git pull
of the skills source failed with "unable to unlink old
'skillshare/SKILL.md': Permission denied".

The sudo child now only receives an internal --replace-binary=<version>
flag that downloads and swaps the binary. The parent waits for it and
does everything else as the invoking user. The needs-password hint shown
by the CLI and dashboard drops its sudo prefix, since a fully root
upgrade reproduces the same breakage.
2026-09-27 02:10:00 +08:00
Willie 5dd83a8412 fix(server): guard metadata store access against concurrent requests
Every API request reloads the skills and agents metadata stores and
swaps the server's pointers under s.mu, but several handlers read or
wrote the stores outside that lock: listing, skill detail, check, the
update stream, and the hash refresh after setting targets. Concurrent
requests could race on the pointer or on the stores' maps, which the
race detector reported hundreds of times under load and which can
crash the server with a concurrent map access.

Store access now always happens under s.mu. Reads return a copy of
the entry so it can be used after unlocking, and target handlers hash
files outside the lock and only write the result under it.
2026-09-26 21:12:57 +08:00
Willie bc01ff1b3e feat(server): let tracked-repo skills take dashboard targets
Setting targets on a skill inside a tracked repo used to be refused,
because writing its SKILL.md leaves the clone dirty and `update` then
skips the repo (or `--force` throws the setting away).

Store those targets in the source's .metadata.json instead, keyed by
the skill's path, and apply them at discovery over the frontmatter.
The clone stays clean, `update` keeps working, and the setting survives
it. Uninstalling a repo or skill drops its entries.

Overrides are saved in the same lock section that writes them: every
API request reloads the store from disk, so saving after unlocking
could lose them, and saving outside the lock raced with other writers.

Refs #295
2026-09-26 21:02:28 +08:00
Willie e259d3c116 fix(plugin): keep a failed host's installed list empty, not null
When an Agent's plugin inventory failed (for example OpenCode with both
opencode.json and opencode.jsonc present), the inventory call overwrote
the host's installed list with nil, which the API encoded as null. The
Plugins page read its length unguarded, so one misconfigured Agent
crashed the whole page.

Host.fail now restores an empty list, covering every inventory path
that reports through it, and the dashboard reads a null list as empty
so a single Agent can no longer take the page down.

Refs #296
2026-09-26 13:11:00 +08:00
Willie 45854c5b57 fix(sync): count manifest-managed hidden entries in target scans
Flat target naming turns a source skill under a dot directory
(e.g. .system/example) into a hidden target entry (.system__example).
Target scans skipped every hidden name, so diff, doctor, status and the
dashboard reported these synced skills as missing, and prune never
removed them after the source skill was deleted.

Hidden entries are now skipped only when the target manifest does not
record them as managed, so files like .skillshare-manifest.json or
.DS_Store are still ignored.

Refs #294
2026-09-25 22:26:51 +08:00
Willie bf32112032 feat(git): show pending remote commits and auto-merge metadata conflicts
Two machines sharing a source repo kept stopping at the same two places.
The Git page had no way to know the remote moved until a push bounced, so
it now fetches once on open and labels Pull with the commit count (status
itself stays offline because the sidebar polls it). And both machines
rewrite .metadata.json on every install or update, so nearly every
diverged pull conflicted there; those conflicts are now merged per entry,
keeping the later installed_at, while any other conflict still stops the
pull and names the files.
2026-09-25 02:22:35 +08:00