* feat(ui): show skill descriptions in the list and let expand all open packages
Reviewing a large library meant opening each detail page just to read
what a skill does (#627, proposal in #631). The resources list response
now carries each skill's and agent's frontmatter description, read in
the same pass as manual-only, and the list shows it truncated under the
name with the full text on hover.
The read now decodes the frontmatter once for both fields. An unquoted
": " in a description is invalid YAML, so a failed decode falls back to
the line scanner instead of dropping the description and manual-only.
Editing a skill refreshes the list too, so the new description shows
right away.
Expand all / Collapse all also opens and closes nested-skill packages
from #624; before, it only touched source and folder groups.
Closes#627
* feat(ui): redesign list packages as an inset panel with a description switch
The chevron-in-a-column nesting read as a second, unrelated table and gave
no room for the descriptions #627 asks for. A count pill after the package
name now opens its skills in one quiet inset panel, with deeper packages
indented under a faint guide, so the hierarchy stays readable at any depth.
The rest closes the gaps left against #627: a Show > Descriptions switch
keeps the compact list one click away, keyboard focus reveals the full
description, a search opens only the packages holding a match without
overwriting what the user left open, and the selection bar says how many
selected items are out of view. Only the name opens the detail page, since
whole-row clicks were easy to trigger by accident once rows held more text.
* fix(ui): keep linked-folder matches when searching a folder group
Searching while grouped by folder took package ancestors only from skills
whose folderOf matched the group key, but groupByFolder keys a linked
folder's skills by the link name. Every skill in a linked folder fell out
of its own group, so the header counted a match and drew no row. The
ancestor pool is now each folder group's unfiltered skills, keyed by
groupByFolder itself, which always contains the group's matches.
* fix(ui): keep the .ss-r .nm.m rule the design token export reads
Merging it into :is(.ss-r, .ss-nr) .nm.m removed the exact selector that
scripts/design/tokens.py looks up for the row name type scale, so the
design token tests failed in CI.
AgentKind.Discover and the skill detail file listing skipped every
hidden directory, including the walk root itself. The `!= "."` guard in
agent.go never fires because walkRoot is an absolute resolved path, so a
sources.agents directory such as ~/.agents discovered zero agents.
Exempt the walk root (`path != walkRoot`), matching the other walks in
the repo. Nested hidden directories are still skipped.
Related to #511 (same class as #514).
A repo installed with `--track --into org` lives at `org/_team`. Only
POST /api/update accepted `org/team`; the dashboard stream and
`update -p` prefixed the whole name (`_org/team`), and global `update`
compared the full name with the basename, so each reported not found.
Move the candidate rule (the name as typed, then `_` added to the last
segment, never leaving the source root) into install.FindTrackedCheckout
and the basename match into install.MatchTrackedRepos, and use them from
the single update, the stream, project and global CLI update. A shorthand
resolves to its own `org/_team`, never to a top-level `_team`; plain names
and full paths keep their behavior.
#474 made install.IsTrackedCheckout (a _-prefixed git repo) the rule for what
a tracked repo is, but several paths still treated any git checkout as one, so
the same directory was classified differently depending on the entry point. An
unprefixed checkout, including a submodule or worktree whose .git is a file,
could be pulled as a tracked repo from the dashboard batch stream, the CLI
--group and project-mode name resolution, and the dashboard single update's
direct-path lookup, and reconcile wrote tracked: true for it.
All of them now use IsTrackedCheckout, so an unprefixed checkout updates as a
regular skill or is skipped, while _-prefixed repos, including --into repos
such as org/_team, still update as tracked repos.
Reconcile no longer registers an unprefixed checkout as tracked or creates an
entry for it from the git remote. An existing entry that an older reconcile
flagged tracked on such a checkout is kept with its source and branch; only the
flag is cleared on the next reconcile, so no user data is removed.
Fixes#476
With --into repos from #472, org/_team and a top-level _team can share a
basename. When org/_team had no entry of its own, the root-skill hash
refresh found _team's entry through GetByPath's basename-only fallback and
overwrote its hashes, which later broke _team's integrity and up-to-date
checks. The refresh now only touches the repo's own entry (exact key, or a
legacy basename key with the matching group).
resolveTrackedRepo built the prefixed --into candidate with filepath.Join,
so on Windows the dashboard result came back as org\_team while #472 made
every other repo name slash-form, and the UI could not match the row.
`install --track --into org` clones to org/_repo, but skill discovery
only marked a skill as in-repo when the first path segment started with
"_". Everything downstream also assumed the repo name was that first
segment. Skills of such a repo were therefore treated as loose skills:
no source in the dashboard, not grouped under the repo, a skill count of
0 in `status`, and an install hint that pointed at the wrong name.
Discovery now records each skill's tracked repo root (RepoRelPath). A
first-segment "_" folder still counts as before; a deeper one counts
only when the walk saw it as a git clone, so a plain org/_drafts folder
stays a folder. The repo's .skillignore is looked up by the same root.
Consumers move from "first segment" to that root:
- the API exposes it as repoPath, and the dashboard groups, updates and
uninstalls repos by it;
- `list` and `status` use it for repo names and counts;
- batch uninstall accepts a nested repo name, so the dashboard can still
remove an --into repo now that its skills are in-repo;
- repo names sent to the dashboard by check and update-all use slash
paths, so they match repoPath on Windows too;
- the install hint suggests `skillshare update org/_repo`.
Fixes#470
Tracked repo metadata is keyed by the repo directory (e.g. "_skills"),
but skills inside the repo are discovered with their full path
("_skills/skills/engineering/tdd"). The lookup found no entry for them,
so /api/resources returned an empty source and the dashboard's update
cards showed "tracked" without owner/repo. Repos whose SKILL.md sits at
the root matched the key directly and looked fine, which is why the
issue appeared to correlate with partial clones.
Nested repo skills now fall back to the repo's tracked entry. They also
ignore a bare-basename match, which GetByPath keeps for legacy short
keys but which can belong to an unrelated top-level skill of the same
name. GetByPath itself is unchanged because update, uninstall and check
depend on its current behavior.
Fixes#467
DELETE /api/repos/{name} now refuses a repo with uncommitted changes or
unreadable git status, and the Dashboard page never sent force, so those
repos could not be removed from it at all.
The two refusals force overrides now carry their own error code. The
page still sends the first request without force; on either code the
confirmation says why and retries with force only when the user
confirms. Other refusals, such as a linked folder that is itself a
skill, keep the plain error because force does not change them.
Refs #442
Review of the shared uninstall operation found that RemoveByNames still
matched a bare key: removing top-level "foo" pruned the legacy entry
"foo" with Group "frontend", which belongs to the installed
frontend/foo. Entries are now matched by their full path only.
Also from review: a dry run no longer runs git status for tracked
repos (it reports no git state), the unused Outcome.Removed is gone,
the two single-item handlers share one status mapping, and the CLI
reports a single target in one place instead of three flag checks.
Refs #442
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
sourceLinks were built before discovery ran, so a target readable at its
root whose subtree failed during traversal stayed available: true with
no warning while sourceLinkWarnings already reported it as incomplete.
Build the items after discovery and treat a recorded skip as unavailable.
A symlink passed as a filepath.Walk root is listed as a single file, so root-of-link skills returned files containing only a dot. Reuse the discovery follow policy to resolve the walk root and compute relative file names from it while retaining logical resource paths.
Cover global/project responses and disabled following, update the existing traversal inventory, and document root-of-link file browsing in all website locales and troubleshooting guidance.
Source links disappeared from the dashboard when discovery found no skills, leaving empty or unavailable folders without an Unlink action. Return first-level links independently with their follow-policy availability and preserve their groups in every skills view, including empty unlink confirmations.
Cover empty and dangling inventory, disabled following, and empty-link removal in list, cards, and tree views; document the API field and rendering.
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.
Share policy-based link identity enrichment between skill list and detail responses. Use that identity in the detail action menu and shared uninstall dialog so followed checkout members keep single-skill confirmation and submit their own flat name.
Prefer exact skill flat names and reject ambiguous basename fallbacks in both uninstall handlers. Route explicit managed repositories before basename lookup so a followed checkout cannot hijack removal of a root skill or repository.
Resolve followed skills with the discovery policy before applying tracked-repo restrictions or prefix-based batch routing. This lets dashboard uninstall move only the selected skill from a followed checkout while retaining managed-repo protection and logical trash restore paths.
Add single and batch handler regressions for global and project modes, skipped links, and tracked-repo members; document the API behavior.
Keep followed checkout updates in the Updates tab and expose unlink as a single quiet icon in group headers. Remove unused checkout metadata and linked-group menus while preserving the confirmation and single-skill uninstall flows.
Match tracked repository affordances without offering whole-repo uninstall for followed user checkouts. Resolve checkout identity for Update repo, keep linked skill uninstall scoped to the selected skill, and explain target preservation, sync impact, and trash recovery before destructive requests.
Expose policy-resolved first-level link identity on listed skills so linked groups can be distinguished from ordinary folders without another request. Add a confirmation that preserves the target folder and explains trash recovery, then refresh the dashboard after unlinking.
Separate user-owned followed Git checkouts into linked_repos in check and resources responses so the Updates tab renders information instead of tracked errors or retry actions. Refuse dashboard updates before Git runs, including force retries and descendant skills, while leaving managed repositories unchanged.
Keep one follow policy per list and status operation and report its skipped links without hiding healthy inventory. Send JSON diagnostics to stderr and expose sourceLinkWarnings in the dashboard resources response, reusing the sync warning formatter.
Use one skills traversal policy for dashboard update discovery, post-pull audits, and staged regular-skill writes so linked repositories are audited and child updates preserve their source links. Keep agent updates and disabled follow behavior on the nil-policy path.
Cover global and project modes through HTTP and SSE handlers, including audit rollback and disabled-follow parity.
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
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
GET /api/audit/{name} joined the client-supplied name to the source
without checking it, so a name such as ../../x scanned a directory
outside the skills source and returned a clean result for it. The
single-skill update, the update stream, and the batch uninstall built
paths the same way; the tracked-repo uninstall route already refused
such names inline.
One helper, validSourceName, now holds that rule: non-empty, relative,
not starting with "..", and no Windows separators or volume prefixes on
any host, as sourcewalk does for declared names. The audit and update
routes answer 400, the update stream reports the name as an error
result, the batch uninstall fails that item, and the repo uninstall
route uses the same helper. Nested names such as group/skill still
resolve.
Closes#402
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
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.
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
The list TUI marks a skill with disable-model-invocation as manual only,
but the dashboard showed that state only inside the skill editor, so a
skill the model never loads looked the same as any other. The skills list,
tiles and the detail page's info column now carry a manual only tag.
The skills API reports manualOnly, read the same way the TUI reads it.
Agents are left out: their harnesses do not read the key.
Refs #283
Make the v0.19.16 sources map actually do something: route all path
reads through Config.EffectiveSkillsSource() / EffectiveAgentsSource() /
EffectiveExtrasSource() so users who set sources.skills / sources.agents /
sources.extras get the configured paths everywhere.
Mechanical changes (~50 files):
- cmd/skillshare/**: replace cfg.Source / cfg.ExtrasSource reads with
cfg.EffectiveSkillsSource() / cfg.EffectiveExtrasSource()
- internal/server/**: replace s.cfg.Source / s.cfg.ExtrasSource reads
with the same helpers; server.skillsSource() / gitignoreDir() use the
global-mode helper
- internal/config/config_validate.go: validate using EffectiveSkillsSource
while still rejecting configs that set neither field explicitly
- internal/config/reconcile.go, registry.go: read via helpers
Behavioral fixes (paired with the migration):
- extras init / extras source / POST /api/extras no longer silently
backfill the legacy extras_source field — the runtime fallback inside
EffectiveExtrasSource() handles it. This stops surprise diffs in users'
config.yaml after running extras commands.
- extras source command writes back to whichever field the user already
configured (sources.extras if set, otherwise legacy extras_source) so
the update is not shadowed.
- source_type in /api/extras and extras list correctly reports "default"
vs "extras_source" by inspecting the raw (unresolved) configured value;
sources.extras counts as configured for this purpose.
UI:
- ui/src/lib/fieldDocs.ts: neutralize sources / sources.* descriptions
so they describe both project and global modes (Phase 5).
Preserved (intentional):
- Direct writes to cfg.Source / cfg.AgentsSource / cfg.ExtrasSource in
init, migrate.go, cloneForSave, and server.reloadConfig — these are
raw-field writes, not path reads.
- Updated test expectations for the removed backfills.
Reorder handleUninstallRepo to move repo to trash before removing the
.gitignore entry, preventing accidental exposure of tracked content
when trashing fails.
ProjectGitignoreTarget now returns empty strings for source paths
outside projectRoot, since parent-relative patterns (../) are invalid
in .gitignore. All six callers guard on empty gitignoreDir and skip
gitignore management for external sources.
reloadConfig now updates s.cfg.Source and s.cfg.AgentsSource from the
refreshed ProjectConfig so project UI handlers see the latest source
directories after a config edit.
Add ProjectGitignoreTarget helper that computes the correct .gitignore
directory and entry prefix based on whether the source path is inside
.skillshare/ or a custom location. All gitignore write and remove
operations (reconcile, install, uninstall, server handlers) now use
this helper instead of hardcoded 'skills/' prefix.
Fix nil projectCfg panics in three server tests that simulate project
mode by setting projectRoot without projectCfg.
Add PATCH /api/resources/{name}/source endpoint to update the source URL
for installed skills. For tracked repos, all sibling skills sharing the
same git remote are updated together with the new repo base URL, and the
git remote origin is also updated.
Frontend changes:
- Add Source URL input field in the skill editor (visible for all skills
with metadata)
- Use shared Input/Textarea components in FrontmatterEditor for visual
consistency, with new size='sm' variant
- Replace custom kind-badge/tracked-badge with shared KindBadge/SourceBadge
components in the editor header
- Preserve original YAML frontmatter formatting when only the body or
source URL is changed (avoids spurious quote style / line wrap diffs)
- Skip diff review dialog when only source URL is modified
Backend fixes:
- Two-pass skill lookup in handleGetSkill: prefer exact FlatName match
over baseName fallback to prevent name collisions between standalone
and tracked repo skills
- Add basename-only fallback in MetadataStore.GetByPath for tracked repo
skills stored with short keys
- Use cfg.Source (global) consistently in save/patch handlers to match
the source used by list/detail handlers in project mode
Closes#142
Backend:
- PATCH /api/targets/:name accepts agent_include/agent_exclude fields
- POST /api/sync-matrix/preview returns agent entries with kind=agent
- GET /api/overview returns agentsSource and extrasSource paths
- Fix ResolveProjectTargets to resolve Agents sub-key (path + filters)
- Add agent filter CLI flags: --add-agent-include, --add-agent-exclude,
--remove-agent-include, --remove-agent-exclude, --agent-mode
- Unify target list JSON/TUI to show agent info and filters
Frontend:
- FilterStudioPage: kind-driven single-context view (?kind=skill|agent)
- KindBadge: use design system tokens (ss-badge + wobble in playful)
- TargetsPage: link to Filter Studio with ?kind= param
- ResourceDetailPage: capitalize type labels (Local, GitHub)
- ResourcesPage: sync activeTab from URL searchParams on navigation
- DashboardPage: show agents/extras source dirs, link skills card
with ?tab=skills for reliable tab switching
- API client: extend updateTarget and previewSyncMatrix for agent filters
Tests:
- Agent include/exclude round-trip persistence
- Invalid agent pattern returns 400
- Preview with agent entries and kind field
- No agent entries for targets without agent path
- CLI target agent filter integration tests
Production code no longer reads or writes .skillshare-meta.json sidecars:
- InstallAgentFromDiscovery: write to centralized store via WriteMetaToStore
- agent_check.go: read from centralized store passed as parameter
- handler_skills.go: read agents from s.agentsStore
- audit.go: readMetaFileHashes reads from parent .metadata.json
- analyzer_metadata.go: extract findMetaEntry to deduplicate walk logic
- uninstall_agents.go: remove entry from centralized store on uninstall
- batchUpdateAgents: pass store to reinstallAgent, remove dead field
- Replace LoadMetadata+nil guard with LoadMetadataOrNew across 6 files
All call sites that looked up metadata entries using filepath.Base()
now use store.GetByPath(relPath) which handles both full-path keys
and legacy basename+group keys transparently.
Also fixes makeInstallOpts() to set SourceDir from updateContext,
preventing nested skill updates from writing metadata to wrong dir.
- Add Tab/Shift+Tab to cycle All/Skills/Agents tabs with count badges
- Tab bar renders at top with bold+underline cyan active, dim inactive
- Tab acts as pre-filter composing with text filter via AND logic
- Cache tab-filtered items in applyFilter() to avoid per-frame recomputation
- Add tabNoun() method to consolidate noun switches across title/filter bar
Agent install --into support:
- Support --into flag for agent install (CLI, server, project mode)
- Support --into=value syntax in addition to --into value
- Server handler_skills uses recursive AgentKind.Discover() for nested agents
- Fix agent metadata read path: use file basename instead of frontmatter name
- Sync --all flag now includes agent sync (not just extras)
Agent content viewer:
- Agents render full-width without sidebar (single .md file, no tree needed)
- Simplified help text and keyboard handling for agent content view
Agent uninstall fixes:
- Use discovery-based resolution instead of hardcoded name+.md path lookup
- Support matching by Name, FlatName, RelPath, or trimmed RelPath
- Compressed confirmation display (list format, max 20 shown)
UI polish:
- Help bar keys highlighted in dim cyan, descriptions in dim gray
- Align 'No items.' text with list title via PaddingLeft(2)
handler_skills list: read agent metadata directly via os.ReadFile
instead of install.ReadMeta (which appends subdirectory). Fixes agents
showing as 'local' instead of 'github'.
handler_skills detail: strip .md from URL param before agent lookup.
Fixes 'skill not found: content-marketer.md' error.
handler_skills: use time.RFC3339 instead of hardcoded format string.
ResourcesPage: use FileText icon for agents, Folder for skills.
handler_skills: skill detail endpoint falls back to agents source when
skill not found. Returns agent .md file content and kind='agent'.
project_reconcile: add ReconcileProjectAgents() to scan agents source,
update registry with kind='agent' entries, and batch-update
.skillshare/.gitignore for agent files and metadata.
TrashPage: add KindBadge next to trashed item names with 'skill'
fallback for old trash items without kind field.
Add agentsSource() helper to Server for mode-aware agent directory
resolution (global: EffectiveAgentsSource, project: .skillshare/agents).
handler_overview: add agentCount field counting .md files in agents dir.
handler_skills: add kind field ("skill"/"agent") to skillItem response.
Support ?kind= query param filter. List agents from agents source dir
alongside skills, with metadata enrichment from per-agent meta files.
handler_targets: add agentPath, agentLinkedCount, agentExpectedCount
fields to targetItem for frontend agent path display.
P2-1: handleUninstallRepo and handleBatchUninstall used s.cfg.Source for
gitignore removal, but project-mode installs write to .skillshare/.gitignore
with a "skills/" prefix. Add gitignoreDir() helper that returns the correct
base directory per mode, and prefix entries in project mode.
P2-2: Registry prune used filepath.Base(repoName) to derive the legacy group,
causing sibling nested repos (e.g., org/_team-skills vs dept/_team-skills) to
collide on the same basename. Now uses exact Group match for nested repos and
restricts legacy basename matching to top-level repos only.
Tests added (15 new):
- resolveTrackedRepo: auto _ prefix, nested, already-prefixed, not-found,
basename fallback
- Registry prune: full-path Group, nested members by prefix, sibling isolation
- Project-mode gitignore: single + batch uninstall
- updateSingle: basename fallback, not-found, regular skill priority
- cleanEmptyParents: multi-level, stops at non-empty, stops at boundary
ReconcileGlobalSkills uses PruneStaleSkills(skillsOnly=false) which
removes agent entries unrelated to the uninstalled repo. The explicit
filter + Save is sufficient; full reconcile deferred to next sync.