DiscoverSourceSkills and DiscoverSourceSkillsLite now load .skillignore
from tracked repos and skip matching paths during walk. This prevents
false warnings in doctor/status/list for vendored SKILL.md files
(e.g. .venv/ inside a tracked repo).
Extract shared skillignore package from internal/install so both
install-time and source discovery use the same filtering logic.
Addresses P1 and P2 from code review on the RWMutex commit:
P1 — Stream handlers now snapshot config under RLock before streaming:
- handleDiffStream: clone Targets map to prevent concurrent map panic
- handleCheckStream, handleAuditStream: snapshot source/projectRoot
- handleUpdateStream: snapshot source (per-item Lock already correct)
P2 — All read handlers now use narrow snapshot instead of defer RUnlock:
- RLock → copy needed fields to locals → RUnlock → do work with locals
- Prevents slow I/O (git fetch, version check, audit scan) from blocking
the auto-reload middleware's exclusive Lock on concurrent requests
Other changes:
- computeTargetDiff takes source parameter (no more s.cfg.Source access)
- Add cloneTargets() helper for safe map copying
- Add writeOpsLogTo/writeAuditLogTo for lock-free oplog writes
The save+reload+error-handling pattern was repeated 8 times across
handler_targets.go (3), handler_extras.go (2), and handler_hub_saved.go
(3). Extract into a single saveAndReloadConfig() method on Server.
One call site in handler_extras.go (extras create) retains separate
save/reload calls because it does MkdirAll between them.
Target mutating handlers (add, remove, update) only called saveConfig()
without the follow-up reloadConfig(), unlike all other mutating handlers
(extras, hub_saved, config). This caused in-memory state to drift from
disk after target include/exclude or mode changes via the UI API.
Add reloadConfig() to all three handlers and tests verifying that
include/exclude filters persist to disk, survive reload, and clear
correctly.
Enable teams to use one project as a shared skills repo where each
developer manages their own targets independently.
- Add `--config local` flag to `init -p` that gitignores config.yaml
- Auto-detect shared repo on clone via .gitignore check (no flag needed)
- Fix partialInitRepair to run interactive prompt instead of auto-select
- Export GitignoreContains, add UpdateGitIgnoreFiles for file-mode entries
- Refactor gitignore helpers to share addGitIgnoreEntries private helper
- Add 5 integration tests covering all new flows
- Add recipe docs page, update init reference, changelogs, release notes
Callers can now distinguish between "no remote configured" and real
failures (e.g. git not in PATH). All existing callers already guard
with err == nil so behavior is unchanged.
dirSize() was called synchronously for every backup version during list
population and on each cursor move, blocking the entire TUI event loop.
- Add ListBackupVersionsLite() that skips dirSize (TotalSize = -1)
- Compute size asynchronously via tea.Cmd, cache in versionSizeCache
- Replace redundant ListBackupVersions() in renderTargetDetail() with
direct os.ReadDir() on the latest backup directory
- Cap per-skill I/O (readSkillDescription + listSkillFiles) to 20 items
in both renderTargetDetail() and renderVersionDetail()
Allow non-technical users to switch branches from the Web UI without
needing a terminal. The dropdown shows local branches first, then
remote-only branches tagged "(remote)".
Backend:
- Add ListLocalBranches, ListRemoteBranches, Checkout to internal/git
- Add GET /api/git/branches (?fetch=true for remote discovery)
- Add POST /api/git/checkout with dirty-check guard + fetch-before-switch
- Oplog instrumentation for checkout operations
Frontend:
- Branch Select dropdown replaces static text when >1 branch exists
- Refresh button to fetch new remote branches on demand
- Dirty state disables switching with visual hint
- Card overflow + z-index fix for playful theme dropdown clipping
Replace the minimal Repo Status Card with a full Repository Info Card
showing remote URL with platform detection (GitHub/GitLab/Bitbucket),
branch tracking info, HEAD commit hash and message, and status badges.
Backend: export GetRemoteURL, add GetHeadMessage and GetTrackingBranch
to internal/git, extend /api/git/status with 4 new fields.
Frontend: new parseRemoteURL utility parses SSH/HTTPS remote URLs and
detects platform by exact hostname match. RepoInfoCard displays parsed
owner/repo with platform icon, clickable link, copy button, and
graceful fallbacks for missing remote or unparseable URLs.
copyDirectoryWithState did not skip .git/ dirs, causing collect to fail
on git-cloned repos (e.g. obra/superpowers) — filepath.Walk would abort
on .git pack files, leaving only empty directories in source.
Add copyDirectorySkipGit() variant and use it in PullSkill. Existing
copyDirectory (used by copy-mode sync) is unchanged.
The update handler used internal/git which ran git commands without
capturing stderr, so failures like divergent branches only showed
"exit status 128". Now runGitWithProgress captures stderr and pipes
it through install.WrapGitError for meaningful messages.
Also adds:
- extractGitFatal() strips verbose hint: lines from git output
- Force checkbox and per-item Force retry button in Update UI
- "(try force update)" / "(try --force)" hint on pull failures
When git commands fail with an empty stderr (common in --progress mode),
wrapGitError() was returning the raw exec.ExitError which only produces
"exit status 128" — giving users no clue about the actual problem.
Now extracts the exit code and maps it to actionable messages:
- exit 128: "repository not found or authentication required"
- exit 128 + token: "authentication token was rejected"
- exit 1: "check network connectivity and repository URL"
Add ability to change sync mode of an extra's target across all three
interfaces: CLI (`extras mode`/shorthand `extras <name> --mode`),
TUI (`M` key with mode picker overlay), and Web UI (inline Select
dropdown per target with PATCH /api/extras/{name}/mode endpoint).
Fix sync behavior so symlinks from a previous mode are auto-replaced
without --force, while regular files still require explicit --force.
Consolidate sync mode list into config.ExtraSyncModes single source
of truth. Fix Select component missing disabled prop and add viewport
collision detection with useLayoutEffect.
Includes 8 integration tests covering mode command happy path, error
cases, shorthand syntax, and mode-switch sync behavior.
- extras remove: replace silent-exit warning with interactive [y/N]
prompt; --force skips the prompt as before
- extras remove: after removing config entry, delete the source
directory if empty and the parent extras/ directory if empty
- sync extras: clean up empty extras/ directory when no extras are
configured (both global and project mode)
- Add removeEmptyDir helper and ExtrasParentDirProject path function
Extras commands had multiple issues with tilde (~) path handling — Go
doesn't expand ~ (that's shell behavior), so paths like ~/.claude/rules
were treated as literal relative paths, creating bogus ~/ directories.
Key fixes:
- Export config.ExpandPath() and call it before os.Stat / sync operations
in extras_list, sync_extras, extras_collect, and doctor
- Unify extras output headers to match target list style (single "Extras"
header instead of per-extra headers, fix double-icon bug in list)
- Add per-target reachability warning in doctor for unreachable extras
targets with expanded path display
Update all 3 E2E runbooks to use extras/<name>/ path layout, add /tmp
cleanup for idempotent re-runs, add migration test step, and create new
comprehensive extras_commands_runbook (18 steps covering init/list/remove/
collect/doctor/tilde/project-mode).
Improve cli-e2e-test skill with 4 new checklist items and 2 rules learned
from test friction (/tmp isolation, symlink write-through, cat>>
idempotency, extras path layout).
Repos like pbakaus/impeccable commit .claude/skills/ alongside their
source skills, causing discoverSkills to find every skill twice (36
instead of 18). Fix by deriving known target dotdir names (.claude,
.cursor, .skillshare, etc.) from targets.yaml project_path values and
skipping them during filepath.Walk.
Non-target hidden dirs (.curated, .system — used by openai/skills) are
still discovered as before. Injection pattern avoids circular import
between config and install packages.
- Move checkExtrasSyncStatus to sync.CheckSyncStatus (shared)
- Add sync.EffectiveMode helper to replace inline mode defaults
- Add config.ValidateExtraMode to centralize mode validation
- Fix sync --all --json outputting two separate JSON objects
by merging extras into the skills JSON output
- Fix swallowed JSON decode error in handleExtrasSync
- Remove dead symlink check in CollectExtraFiles (filepath.Walk
follows symlinks so the check was unreachable)
- Fix integration test extras paths to use extras/<name>/ layout
- GET /api/extras — list extras with sync status
- GET /api/extras/diff — per-file diff for extras
- POST /api/extras — create new extra
- POST /api/extras/sync — sync extras (all or by name)
- DELETE /api/extras/{name} — remove extra from config
ValidateExtraName rejects empty names, reserved words ("skills", "extras"),
and names that don't start with alphanumeric or contain invalid characters
(slashes, dots, spaces, leading dash/underscore). ValidateExtraNameUnique
prevents duplicate extra names in config.
JihuLab (jihulab.com) is a GitLab distribution that supports nested
subgroups but its hostname doesn't contain "gitlab". Add "jihulab" to
the built-in host detection so URLs like jihulab.com/group/sub/project
are parsed as full-path repos without requiring gitlab_hosts config.
- TestParseOpts_ProjectMode_NoFallbackToGlobal: server project mode
returns empty hosts when project has none, never falls back to global
- TestParseOpts_ProjectMode_MergesEnvVar: env var merges with project
config, not global
- TestLoadProject_GitLabHosts_EffectiveGitLabHosts: project config
EffectiveGitLabHosts() merges env var while GitLabHosts stays
config-file-only
Move SKILLSHARE_GITLAB_HOSTS merge from Load()/LoadProject() to new
EffectiveGitLabHosts() methods so the GitLabHosts field stays config-file-only
and Save() never writes env-only hosts to disk.
Also fix project-mode direct install dropping gitlab_hosts from the
synthetic Config, and server project-mode parseOpts() falling back to
global config when project had no hosts configured.
Support CI/CD pipelines that need gitlab_hosts without a config file.
The env var is comma-separated and merged (deduplicated) with config
file values in both global and project modes.
Also cleans up from review:
- Extract isValidGitLabHostname to share validation logic
- Remove redundant ToLower before EqualFold in isGitLabHost
- Revert 3 no-opts call sites back to ParseSource()
- Add parseOptsFromProjectConfig helper to reduce inline duplication
- Fix project mode LoadProject missing env var merge
- Add install.md tip about custom GitLab domains
Self-managed GitLab instances on custom domains (e.g., git.company.com)
without "gitlab" in the hostname were parsed with 2-segment owner/repo
split instead of full-path repo. Users can now declare extra GitLab
hostnames in config so the parser treats them like gitlab.com (nested
subgroup support).
- Add `gitlab_hosts` field to both global and project Config structs
- Validate entries at load time (reject scheme, path, port, empty)
- Add `ParseOptions` / `ParseSourceWithOptions` to thread hosts into parser
- Add `isGitLabHost()` helper combining built-in + custom host detection
- Extend `InstallContext` interface with `GitLabHosts()` method
- Update all ParseSource call sites (install, search, update, server)
- Add JSON schemas for both config files
- Add unit tests for config validation and source parsing
- Add E2E runbook for devcontainer verification
- Document in configuration reference and URL formats appendix
The initial GitLab subgroup fix treated all multi-segment HTTPS paths as
full repo paths, regressing GHE and generic host subdir installs like
github.mycompany.com/org/repo/path/to/skill.
Add host-based heuristic: hosts containing "gitlab" use full-path-as-repo
behavior; all other hosts retain the original 2-segment owner/repo split
with remainder as subdir. This preserves backward compatibility for GHE,
Gitea, Gogs, and other non-GitLab platforms while correctly handling
GitLab nested subgroups.
Replace rigid 2-segment HTTPS regex with flexible path parsing that
treats the entire URL path as a repo when no explicit boundary exists.
Use .git suffix as the canonical repo/subdir delimiter, preserving
/-/ (GitLab) and /src/ (Bitbucket) web URL markers.
This fixes install failures for URLs like
onprem.gitlab.internal/org-group/sub1/sub2/project which previously
misinterpreted extra path segments as monorepo subdirectories.
Increment() now clamps p.current to p.total so concurrent callers
cannot produce a "11/10" display. RenderInlineBar() returns a safe
empty bar when total <= 0 instead of panicking on division by zero.
Phase transitions during batch update (e.g. pull repos → update skills) were
printing separate PhaseHeader lines that created a second visible progress bar.
Now phases use ProgressBar.SetHeader() which overwrites the header line in place
via ANSI cursor-up, keeping a single continuous bar from start to finish.
colorSkillPath() and colorSkillPathBold() used Color("241")~Color("249")
gradient for subdirectory segments — fixed palette grays that render
inconsistently across terminal themes. Replace with tc.Dim (SGR faint).
\033[90m (bright black) renders too dark on many terminals and themes.
Replace with \x1b[0;2m (SGR dim attribute) which adapts to any foreground.
Changes:
- Add Dim constant and DimText() helper in internal/ui
- Replace all pterm.Gray() calls with DimText() in pterm.go
- Replace all ui.Gray references with ui.Dim across cmd/skillshare/
- Replace fixed 256-color grays (Color("247"/"242")) with lipgloss
.Faint(true) in TUI styles (tui_colors.go, list_tui.go, log_tui.go)
- Fix progress bar layout: put count after % at fixed position,
title last (truncated) — prevents count from being pushed off-screen
- Extract duplicate renderBar() from diff.go and search_batch.go
into shared ui.RenderInlineBar() helper
Replace pterm's ProgressbarPrinter with a custom implementation using
\r overwrite for flicker-free rendering. The new bar uses block chars
(■/・), orange color with dim text for labels, and dynamic title width
based on terminal size.
Key improvements:
- Render throttling (50ms) prevents I/O overload on large skill counts
- Git progress messages auto-normalized in UpdateTitle
- Final 100% frame preserved on screen after Stop()
- Cursor hidden during progress, restored on completion
The detail panel header was rendered on L00 (same row as the list title),
making the skill name invisible. Push content to L01 with a leading newline
and use colorSkillPath for consistent group/name coloring matching the left
panel. Compress status/date/targets into a single metadata line to save
vertical space. Show full source URL and synced targets in Quick Facts.
Also suppress devcontainer stderr redirect message in --json/--format modes
and fix audit summary field name (analyzerCount → activeAnalyzers).
Keep sync dry-run diagnostics on stdout by default to avoid non-JSON behavior regressions. Introduce sync.DiagOutput so JSON mode can silence diagnostics via io.Discard at command entry and restore after execution. Update search JSON integration assertion to match fully silent progress behavior.
- Extract writeJSONResult() to replace 5 identical write+silent-error trailers
- Replace local isRepoDirty() with existing git.IsDirty() (8 call sites)
- Use defer for suppressUIToDevnull restore instead of manual calls at each
return site in install.go and update.go (prevents stdout leak on new paths)
- Fix ResolvePolicy always filling EnabledAnalyzers, which broke the
nil-means-all contract for ForPolicy; add EffectiveAnalyzers() for display
- Simplify collectOutputJSON double map allocation
- Remove unnecessary string conversion in writeJSONError
When no explicit analyzer filter was configured, status --json reported
an empty analyzers array. Now ResolvePolicy backfills the full list of
built-in analyzer IDs so the JSON output reflects what actually runs.
Add Registry.IDs() helper with deduplication (markdownLinkAnalyzer and
staticAnalyzer share the "static" ID).
Address code review feedback from PR #67:
- Move all dry-run diagnostic messages (internal/sync) from stdout to
stderr so they don't pollute JSON output
- Suppress UI output (spinners, headers, warnings) in uninstall --json
mode by adding jsonOutput guards across all phases
- Return nil after writeJSONError to prevent dual output (JSON + plain
text error) in install and sync commands
- Handle json.MarshalIndent error in writeJSONError with stderr fallback
- Merge duplicated discovery logic (JSON vs non-JSON branches) in sync,
sync_project, diff, and collect using nullable spinner pattern
- Remove checkRepoDirty alias in status.go, call isRepoDirty directly
- Remove extractJSON test workaround — tests now assert stdout is valid
JSON directly, enforcing the clean-stdout contract
Detect inline API keys, tokens, and passwords in skill files via 10 new
hardcoded-secret-* rules (HIGH severity): Google API keys, AWS access
keys, GitHub PATs (classic + fine-grained), Slack tokens, OpenAI keys,
Anthropic keys, Stripe keys, PEM private key blocks, and generic
api_key/password assignments.
Closes a gap where GitHub Secret Scanning caught a Google API key in a
skill file but our audit engine missed it entirely.
Replace sequential POST loops and blocking useQuery calls with
Server-Sent Events for real-time progress feedback on Update, Check,
and Diff operations. Extract shared helpers to reduce duplication:
- initSSE() in middleware.go consolidates SSE preamble across 4 handlers
- createSSEStream() in client.ts unifies EventSource lifecycle management
- StreamProgressBar component replaces 3 near-identical progress bars
- handleDiff now delegates to computeTargetDiff (eliminates ~150 lines)
- Paginate audit results (20 cards + 10 findings per card) to prevent
DOM overload with large skill counts
Add metadataAnalyzer (skill scope) that cross-references SKILL.md
metadata against the actual git source from .skillshare-meta.json:
- publisher-mismatch (HIGH): description claims an org but repo owner
differs (e.g. "from Acme Corp" but sourced from evil-fork)
- authority-language (MEDIUM): uses "official"/"verified" language but
source is not a well-known organization
Both rules are disable-able via audit-rules.yaml. The analyzer reads
meta from disk (zero API signature changes) and avoids import cycles
by defining its own minimal metaJSON struct.
When git operations fail due to SSL certificate issues (self-signed certs,
internal CAs), the error message now suggests GIT_SSL_CAINFO, GIT_SSL_NO_VERIFY,
and SSH as alternatives instead of showing raw git stderr.
Env vars like GIT_SSL_CAINFO are already inherited via os.Environ() in
gitCommand() — the actual gap was error detection and user guidance.
Add server-sent events endpoint for audit scans so the web UI can
display per-skill progress instead of waiting for the full batch.
Backend:
- Extract discoverAuditSkills and processAuditResults from handler
- Add handler_audit_stream.go with SSE streaming endpoint
- Register GET /api/audit/stream route with read-lock middleware
- Add CORS headers for SSE connections in middleware
Frontend:
- Add AuditStreamPanel and AuditResultCard components
- Wire streaming audit into AuditPage with start/stop controls
- Add useAuditStream hook and streamAudit API client method
- Add per-skill audit action on SkillDetailPage
- Add check-based update detection on UpdatePage
- Configure Vite proxy for SSE endpoint