Remove rules that generated excessive false positives:
- source-repository-link (HIGH): legitimate docs links flagged too often;
external-link (LOW) already covers the same URLs
- env-access (MEDIUM): process.env references are ubiquitous in JS skills
- obfuscation-1 (long base64): caught README badges and binary data
Recalibrate credential-access severity levels to reduce alert fatigue:
- etc-passwd/sudoers: HIGH → MEDIUM (read-only system files, not secrets)
- shell-history, openvpn: MEDIUM → LOW (low-impact reconnaissance)
- auth-log: LOW → INFO (log files, no direct credential exposure)
Update all affected unit tests, integration tests, red team assertions,
skill references, and website docs to match.
P1: mergeYAMLRules Phase 2 used full-object replacement for overlays
without Enabled set, causing severity-only entries ({id, severity}) to
wipe regex/pattern/message from the base rule. compileRules then failed
with "empty regex". Now uses field-level merge: only entries with Regex
do full replacement; otherwise Enabled and Severity are patched
individually.
P2: TogglePattern enable only removed pattern-level disable entries,
leaving ID-level {id: "x", enabled: false} untouched — both builtin
and custom rules. Now clearDisableForPattern matches by builtin ID set
OR by r.Pattern == pattern, covering custom rules in the same group.
Implement a full bubbletea TUI with two navigation levels:
- Level 1 (patterns): shows pattern categories with rule count, disabled
count, and max severity; press Enter to drill in, d to toggle pattern
- Level 2 (rules): shows individual rules with severity and status; detail
panel shows full ID, regex, message, source; d toggles single rule,
D toggles entire pattern, Esc returns to Level 1
Also adds audit.ResetGlobalCache() to clear cached rules after YAML
mutations, enabling live-reload within the TUI session.
Introduce rules_list.go with exported types and functions for listing
all audit rules with their enable/disable status and source layer
(builtin/global/project). Includes PatternSummary for grouped stats
and UniquePatterns for pattern enumeration. Disabled rules appear in
listings with Enabled=false so UIs can display toggle state.
Extend mergeYAMLRules to handle overlay entries that specify a pattern
but no id — these apply to all base rules sharing that pattern name.
Processing uses two phases:
1. Pattern-level overlays (bulk disable / severity override)
2. ID-level overlays (can override pattern-level, e.g. re-enable
an individual rule after a pattern-level disable)
An explicit enabled:true in an id-level overlay clears a disabled flag
set by a pattern-level overlay.
Strengthen the audit engine with new detection capabilities inspired by
skilld's sanitization research and Snyk ToxicSkills findings:
- Expand prompt-injection-1 to detect OVERRIDE:/IGNORE:/ADMIN:/ROOT: prefixes
- Add prompt-injection-2: agent directive XML tags (<system>, <instructions>, etc.)
- Add prompt-injection-3: DEVELOPER MODE, JAILBREAK, DAN MODE directives
- Add data-exfiltration-2: external images with query params (exfil vector)
- Add credential-access-3/4: /etc/shadow, /etc/passwd and related system files
- Add TierInterpreter (T6): classify python/node/ruby/perl/lua/php as
Turing-complete runtimes separate from build tools
- Add tier-interpreter and tier-interpreter-network combination findings
- Add cross-skill-cred-interpreter rule for credential × interpreter pairs
- Update docs to reflect 38 rules and new tier/cross-skill patterns
pterm's cursor package writes hide/show ANSI codes (\x1b[?25l/h)
directly to os.Stdout, bypassing WithWriter(). The previous fix
(SetProgressWriter to stderr) was insufficient because pterm still
created progress bar objects that triggered cursor.Hide()/Show().
Add SuppressProgressToStderr() which both redirects ProgressWriter
and forces isProgressTTY() to return false, preventing pterm
components from being instantiated at all. Extract assertNoANSI
test helper and cover all three structured formats (JSON, SARIF,
Markdown).
Also update pre-commit E2E runbook with steps 10-11 (framework
integration tests) and add pre-commit-hook recipe to sidebar.
Replace nested pair-checking loop with set-based approach:
- Single pass builds capability sets (credReaders, netSenders, etc.)
- Each rule checks non-empty sets → one summary finding per rule
- 100k skills: ~15ms vs ~30s+ (benchmark included)
- Use DiscoverSourceSkillsLite for audit discovery (skip frontmatter)
- Use patternDataflowTaint constant instead of raw string
P1: SARIF startLine=0 is invalid per spec — GitHub Code Scanning rejects
it. Use *sarifRegion pointer with omitempty so findings without line
numbers (tier, cross-skill, analyzability, content-integrity) produce
valid SARIF output without a region field.
P2: Server audit endpoint (handleAuditAll) now calls CrossSkillAnalysis
after collecting results, matching CLI behavior. Previously the API
returned only per-skill findings without cross-skill interaction checks.
Also guard fmt.Println() in auditFiltered() for structured output
consistency (same fix as auditInstalled in previous commit).
- Add --format markdown for self-contained audit reports (GitHub Issues/PRs)
- Simplify pterm box rendering: replace DefaultBox with separator-line style
- Remove unused SkillBox, SummaryBox, SkillDisplay, wrapText helpers
- Color Box titles cyan (consistent with HeaderBox)
- Add breathing room: blank lines around progress bar and after summary
- Show "skillshare audit (project)" header title in project mode
- Append elapsed time to cross-skill synthetic results to fix slice mismatch
Two individually harmless skills can form an attack chain when installed
together (e.g., one reads credentials, another has network access).
After per-skill scans complete, the audit engine now extracts a capability
profile from each Result and checks for dangerous cross-skill combinations:
- cross-skill-exfiltration (HIGH): credential reader + network sender
- cross-skill-privilege-network (MEDIUM): privilege commands + network
- cross-skill-stealth (HIGH): stealth skill alongside high-risk skill
Rules only fire for complementary pairs (each skill lacks the other's
capability) to avoid duplicating per-skill findings. Results appear as
synthetic "_cross-skill" entry in all output formats (text, JSON, SARIF, TUI).
Adds forward taint propagation analysis that detects multi-line data
exfiltration patterns where sensitive data flows through variables or
temp files before reaching a network sink.
Tracks three taint sources: credential file reads (~/.ssh, .env, etc.),
sensitive env var references ($TOKEN, $API_KEY, etc.), and command
substitution outputs. Propagates taint through variable assignments,
pipe chains, and temp file redirections. Generates HIGH-severity
dataflow-taint findings when tainted data reaches network commands.
- Add ScanShellDataflow for shell scripts and ScanMarkdownDataflow
for fenced code blocks in markdown
- Add DeduplicateDataflow to avoid duplicate findings with existing
per-line pattern rules
- Integrate into scanSkillImpl and ScanFileWithRules scan paths
- Can be disabled via audit-rules.yaml (id: dataflow-taint)
Classify every shell command in skill files into 6 behavioral tiers
(read-only, mutating, destructive, network, privilege, stealth) as an
orthogonal dimension to pattern-based severity. This lets audit report
"what kind of actions" a skill performs, not just "which known patterns
it matches."
- Add tier engine with 80+ command mappings and stealth regex patterns
- Detect tiers in scripts and Markdown fenced code blocks separately
- Generate combination findings: T2+T3→HIGH, T5→CRITICAL, T3>5→MEDIUM
- Display tier profile in CLI text, TUI detail panel, and JSON/API output
- Exclude .skillshare-meta.json from totalBytes to fix analyzability ratio
- Document tier definitions, profile output, and combination findings
Analyzability score measures what percentage of a skill's content is
auditable plaintext vs binary/non-scannable. Surfaces in CLI text
output, JSON, SARIF, and Web UI API responses.
- Add TotalBytes, AuditableBytes, Analyzability fields to audit.Result
- Emit INFO-level low-analyzability finding when score < 70%
- Show "Auditable: X%" in per-skill and summary text output
- Expose analyzability in server API (auditResultResponse + auditSummary)
- Add .pre-commit-hooks.yaml for pre-commit framework integration
- Add pre-commit hook recipe docs and audit.md cross-references
Previously --format json/sarif completely suppressed spinner and progress
bar, making the CLI appear frozen during long scans.
Now spinner/progress output is redirected to stderr via ProgressWriter,
so progress indicators remain visible while structured data stays clean
on stdout. This follows the Unix convention: data → stdout, diagnostics
→ stderr.
Also:
- Extract fileTTY() shared helper (DRY IsTTY + isProgressTTY)
- Add format constants (formatText/JSON/SARIF) and isStructured()
- Capture/restore previous ProgressWriter in defer (defensive)
Add ToSARIF() function that converts audit Results into SARIF 2.1.0
format for CI/CD integration (GitHub Code Scanning, etc.).
- 12 structs modeling the SARIF schema subset we need
- Severity mapping: CRITICAL/HIGH→error, MEDIUM→warning, LOW/INFO→note
- Rule deduplication by pattern (first occurrence wins)
- SARIFOptions for tool version and base URI
- 4 test cases: empty, findings, severity mapping, JSON round-trip
filepath.Walk/WalkDir silently skip symlinked root directories because
os.Lstat sees the root as a non-directory symlink and never enters it.
This affected update, uninstall, reconcile, and server handlers when
the source directory was a symlink (common with dotfiles managers).
Extract ResolveSymlink() from sync.go into internal/utils/path.go and
apply it at every filepath.Walk/WalkDir callsite on source/target dirs.
Also add containment guards to resolveGroupSkills (uninstall) and
resolveGroupUpdatable (update) to reject group directories that are
symlinks pointing outside the source tree.
Upgrade prompts ("Install built-in skill?" and "Upgrade to vX?")
left their input lines visible after the user responded, breaking
the tree-drawing structure. Add ui.ClearLines() to erase prompt
lines via ANSI cursor-up + clear-line, so StepEnd connects cleanly
to the preceding tree node. No-op on non-TTY.
When source or target directories are symlinks managed by external tools
(e.g., dotfiles managers), sync would incorrectly delete them during
mode conversion. This change:
- Add resolveWalkRoot() so filepath.Walk follows symlinked source dirs
- Add isSymlinkToSource() with canonicalization to distinguish internal
symlink-mode targets from external symlinks (alias-safe)
- Protect external target symlinks in both merge and copy mode
- Allow FindLocalSkills to scan through external symlinked targets
- Handle external symlinks in CheckStatusMerge as normal directories
When a target directory is an external symlink (e.g., dotfiles manager
pointing ~/.claude/skills/ elsewhere), sync previously deleted it
unconditionally during symlink→merge/copy mode conversion. This also
affected DiscoverSourceSkills which failed to traverse symlinked source
directories because filepath.Walk uses Lstat on root.
Changes:
- Add resolveWalkRoot() helper; use in both Discover functions so
filepath.Walk enters symlinked source directories
- Add isSymlinkToSource() to distinguish source-pointing symlinks
(symlink mode residue, safe to remove) from external symlinks
- Pass explicit sourcePath to SyncTargetMergeWithSkills and
SyncTargetCopyWithSkills for reliable symlink origin detection
- Fix FindLocalSkills to use utils.ResolveLinkTarget (handles relative
symlinks correctly) and follow external symlinks for scanning
- Fix CheckStatusMerge to follow external symlinks instead of returning
StatusConflict
- Add 17 unit tests and 6 integration tests covering symlinked source,
target, both, chained symlinks, idempotency, prune, and pull
Homebrew formula updates lag GitHub Releases by hours/days, causing
Homebrew users to see "update available" when `brew upgrade` says
already up-to-date. Now detect install method from executable path
and query `brew info --json=v2` for Homebrew installs instead of
the GitHub Release API.
- Add InstallMethod type with DetectInstallMethod() and
FetchBrewLatestVersion() in internal/version/homebrew.go
- Check() accepts InstallMethod to branch fetch strategy
- UpdateNotification() accepts upgradeCmd (shows correct command)
- UpgradeCommand() method on InstallMethod centralizes the mapping
- doctor.go reuses versioncheck.Check() (fixes semver string
comparison bug in old fetchUpdateMessage)
- Remove duplicated isHomebrewInstall() from upgrade.go
- Skip post-command version check for doctor (has its own)
- Remove duplicated RateLimitError, newGitHubRequest, newGitHubClient,
checkRateLimit from internal/version — now uses internal/github client,
which also fixes missing `gh auth token` fallback for version checks
- Add GH_TOKEN as fallback in install.resolveToken() for GitHub platform,
matching github.GetToken() resolution order (GITHUB_TOKEN → GH_TOKEN)
- Add GH_TOKEN to sanitizeTokens() redaction list
- Extract shared IsAuthError() in install/auth.go; push.go and
wrapGitError() both delegate to it (adds "Access denied" coverage)
- Consolidate shouldFallback*TrackedClone auth checks via isAuthOrAccessError()
- Thread single *http.Client through search.Search() and all callees,
reducing ~60 redundant TLS handshakes to connection-pooled reuse
- Pass *http.Client to downloadFile() instead of creating per-file clients
- Replace sourceHost() with extractHost() (same package, identical purpose)
- Remove unused isGitRepo() alias; callers use IsGitRepo() directly
- Extract shared applyDetailScroll into tui_helpers.go (was duplicated
across diff_tui.go, log_tui.go, audit_tui.go)
- Create utils.FileHash and utils.FileHashFormatted, replacing private
copies in install/meta.go and audit/audit.go
- Export install.MetaFileName constant, use from diff_file.go
- Replace double SHA-256 in fileContentsEqual with streaming byte compare
for early-exit efficiency
- Add 1MB size cap to generateUnifiedDiff to prevent performance issues
- Move noTUI into diffRenderOpts struct, simplifying function signatures
- Add expand state (Enter key toggle) for file-level diff display
- Show file list with +/-/~ icons for add/delete/modify files
- Show colored unified diff (green/red/cyan) for modified files
- Add source/target mtime rows in detail header
- Replace "Actions" section with "Next Steps" using prominent styling
- Update list item description with sync/local breakdown
- Add stat counts (err/diff/ok) to TUI title bar
- Fix freeze: defer file walk to Enter toggle instead of every render
- Fix gofmt in trash_tui.go and restore.go
Add ListBackupVersions() and ListTargetsWithBackups() for TUI timestamp
selection. Fix merge-mode targets producing empty backups by following
top-level symlinks in copyDirFollowTopSymlinks().
Include skill description from SKILL.md frontmatter in the discover API
response, add the optional description field to the DiscoveredSkill
TypeScript interface, and render it inline in the SkillPickerModal.
Filter search also matches against descriptions. The --into directory
input already existed in InstallForm, so no change was needed there.
handleDiscover and handleInstallBatch now use DiscoverFromGitSubdir
when the source URL contains a subdirectory path, enabling the skill
picker modal to appear for multi-skill subdirectories.
Fix SkillPickerModal content disappearing on checkbox click caused by
Card's overflow-hidden creating a scroll container — browser's
scrollIntoView on the sr-only input would scroll content out of view.
Use overflow-clip instead to preserve border-radius clipping without
creating a scroll context.
Issue #56 (workaround)
Remove 2-space indent from spinner Success/Fail/Warn and
SuccessMsg/ErrorMsg so they align with other ui.Success output.
Move backup after discovery in sync so it is skipped when there
are 0 skills to sync — no point backing up if nothing will change.
- Fix non-TTY progress bar not incrementing for grouped skills in Phase 2
- Fix Unicode truncation in UpdateTitle using runewidth instead of len()
- Fix IP address regex missing end-of-line bare IPs (add $ anchor)
- Fix gofmt alignment in updateTarget struct comments
- Extract sliceHasName[T] generic helper to deduplicate universal detection
- Add nil meta defensive warning with skipped count in grouped updates
- Replace hand-rolled contains/containsStr with strings.Contains in tests
- Add red team TC-20b for inline code fetch-with-pipe detection
Optimize grouped repo updates by skipping reinstall when installed metadata already matches remote state (commit match, or subdir tree-hash match when HEAD moved).
Emit per-skill grouped progress events so the batch progress bar advances continuously instead of appearing stuck at 0%.
When skipping by tree-hash match, best-effort refresh meta.Version to latest commit to keep status/version output current.
Add unit tests covering commit/tree-hash skip behavior and metadata version refresh.
When update --all processes many skills from the same repo, avoid reinstall/copy when installed metadata already matches the fetched repo state.
For root skills, skip when meta.Version matches repo commit. For subdir skills, require both meta.Version and meta.TreeHash to match. Added unit tests for match/mismatch cases.
- Add PhaseHeader UI helper for "[1/3] Pulling N tracked repos..." labels
- Show phase headers between batch update phases when 2+ phases exist
- Skip phase headers when only 1 phase (avoids meaningless "[1/1]")
- Lazily start progress bar after first phase header to fix ordering
- Add scanning spinner during filepath.Walk for large skill sets
- Cache metadata in updateTarget to avoid redundant ReadMeta calls
Size was computed by walking every file in every local skill directory
during the scan phase, before the user even confirms. CLI never displays
the size. Move to lazy on-demand calculation in the Web UI handler only.
UpdateGitIgnore/RemoveFromGitIgnore were called per-skill inside loops,
re-reading the entire .gitignore each time. With large files (100K+ lines
in project mode), this caused O(N×M) complexity and process hangs.
- Add UpdateGitIgnoreBatch and RemoveFromGitIgnoreBatch (single read/write)
- ReconcileProjectSkills: collect entries during walk, batch-update once
- uninstall (global + project): batch-remove after loop, not per-iteration
- Fix spinner WithRemoveWhenDone in StartSpinnerWithSteps
- Add E2E runbook for gitignore batch performance validation
Add generic OperationSummary() with Metric struct for consistent
batch output. Refactor SyncSummary/UpdateSummary as thin wrappers.
Apply unified patterns to collect, uninstall, backup:
- Spinner for scanning phases
- StepDone/StepSkip/StepFail semantic indicators
- OperationSummary for batch results
- Header before collect scanning for context
- Spacing before backup "Recent backups" section
Fix pterm spinner RemoveWhenDone bug: Success/Fail/Warn messages
were silently eaten because Stop() clears the line after printing.
Use Stop() + lightweight symbols (✓/✗/!) instead of pterm badges.
Also replace SuccessMsg/ErrorMsg pterm.Success/Error badges.
Add 5 new audit rules (31 → 36 total) to close detection gaps:
- fetch-with-pipe (HIGH ×3): curl/wget piped to sh/bash/python/node/etc.
Previously only double-pipe (||) was caught by shell-chain; single
pipe (|) to interpreters — the classic RCE vector — was undetected.
- ip-address-url (MEDIUM): URLs with raw IP addresses that bypass DNS
security controls; excludes RFC 1918 and loopback ranges.
- data-uri (HIGH): data: URIs in markdown links that may embed
executable content.
fetch-with-pipe is suppressed inside markdown code fences (tutorial
context), matching existing behavior for destructive-commands et al.
Includes unit tests (25 cases), code fence suppression test, red team
TC-18 through TC-21, and audit.md documentation updates.
The universal target pointed to ~/.config/agents/skills (wrong) instead
of ~/.agents/skills, causing `npx skills list` to miss synced skills.
Also auto-recommend universal as a target whenever any CLI is detected
during `init` and `init --discover`, since ~/.agents/ won't exist on
disk until skillshare creates it.
Closes#54