The 1294-line audit.md mixed command reference, conceptual teaching,
rule management, and CI/CD recipes. Split into:
- understand/audit-engine.md — threat model, detection rules, risk
scoring, command tiering, cross-skill analysis
- reference/commands/audit-rules.md — rule management, custom rules,
built-in rule IDs table (adds 4 cross-skill finding IDs)
- how-to/recipes/ci-cd-skill-validation.md — absorb output format
docs (SARIF, Markdown, JSON+jq, exit codes)
- reference/commands/audit.md — slim to ~370 lines, link to new pages
Also fixes:
- Replace ghost rule `system-writes-0` with `insecure-http-0` in all
docs (the old ID no longer exists in rules.yaml)
- Update 3 cross-page links in security.md
- Update 1 rule ID in configuration.md
- Add sidebar entries for both new pages
applyPolicyToSummary was called in cmdAudit after auditInstalled/
auditFiltered returned, but both functions call presentAuditResults
(which launches the TUI) internally — so summary.PolicyProfile was
empty when the TUI rendered. Move applyPolicyToSummary before
presentAuditResults in both functions.
Also fix dedupe default label (global is now the default, not legacy)
and update audit/status docs to reflect current output format.
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.
Update CHANGELOG.md and website changelog with v0.16.9 entry covering
interpreter tier (T6), expanded prompt injection/credential access
detection, and regex bypass fixes. Sync audit.md with new rule IDs
and detection descriptions. Add contributor reneleonhardt to README.
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
- 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).
The analyzability score feature (committed in ae7bd21) was missing
dedicated documentation. Also backfills content-oversize (from 3ad814d)
which was never added to the docs.
- Add "Analyzability Score" section explaining score meaning, threshold,
exclusions, and CLI/JSON output format
- Add content-oversize to Content Integrity table
- Add low-analyzability to LOW/INFO severity listing
- Add low-analyzability (INFO) and content-oversize (MEDIUM) to
Built-in Rule IDs table
- Update :::note block to mention new structural/analyzability checks
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
- Add --format flag to Options table, mark --json as deprecated
- Add SARIF output section with severity mapping table
- Add GitHub Actions Code Scanning example with SARIF upload
- Update usage examples and recommended workflow
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.
Covers 13 test scenarios for dotfiles manager symlink support:
symlinked source/target, chained symlinks, merge/copy modes,
status/diff/list/collect through symlinks, and idempotency.
Update website diff page and built-in skill reference to reflect:
- Interactive TUI with left-right panel layout (default on TTY)
- New flags: --stat, --patch, --no-tui
- Redesigned category labels (New, Modified, Local Override, etc.)
- File-level diff examples
Also remove unused results field from diffTUIModel and replace
custom countOccurrences with strings.Count in tests.
Support glob patterns (*, ?, [...]) in skill name arguments across all
three commands. Matching is case-insensitive.
- install: --skill and --exclude flags support glob (e.g. -s "core-*")
- update: positional names support glob (e.g. update "core-*")
- uninstall: positional names support glob (e.g. uninstall "core-*")
Shared helpers isGlobPattern() and matchGlob() extracted to helpers.go
to avoid duplication. Includes 25 new tests and updated docs.
Add bubbletea-based TUI with multi-select, fuzzy filter, and inline
restore/delete/empty operations. Follows the same dispatch pattern as
list and log: JSON → TUI (TTY + items + !--no-tui) → plain text.
- Change universal description from "(recommended)" to "(shared agent directory)"
in detect, checklist, and discover flows
- Add FAQ section on using universal alongside npx skills: merge vs copy mode
differences, name collision caveats, lock file visibility
- Add universal info admonition to supported-targets page
- Add E2E runbook for universal target path verification
- Update CHANGELOG and release notes for v0.16.4
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
- Add Duplicate Detection section to install.md (same-repo skip,
cross-path detection, conflict with different repo)
- Update --force flag description to mention dedup override
- Add v0.16.4 changelog entry to CHANGELOG.md and website
Add Verdent (verdent.ai) as a built-in target with standard skill paths.
Update target count from 49+ to 50+ across README, homepage, and docs.
Sync changelog and release notes with the new target entry.
- Add bubbletea TUI for multi-skill audit results (side-by-side layout
with skill list + detail panel, severity sorting, fuzzy filter)
- Add progress bar during parallel scan via onDone callback
- Add --quiet/-q flag to suppress clean skill lines
- Add --yes/-y flag to skip large-scan confirmation prompt (>1000 skills)
- Add --no-tui flag to force plain text output
- Scale parallel workers to NumCPU (bounded 2–32) instead of fixed 8
- Use file cache in content-integrity check to avoid redundant disk reads
- Replace line splitting with streaming scan (avoid full []string alloc)
- Remove snippet truncation (preserve full trimmed line)
- Differentiate LOW (bright blue) and INFO (medium gray) severity colors
- Centralize log TUI color styles into `lc` struct
- Export DisplayWidth and add BoxWithMinWidth/HeaderBoxWithMinWidth for
aligned header + summary box widths
- Update docs with TUI controls, new flags, and CI/CD auto-skip behavior
When multiple targets produce the same diff output (e.g. all missing
the same skills), they are now merged into a single block with a
combined header like "claude, agents" instead of repeating the full
list for each target. Fully synced targets are also consolidated
into one line (e.g. "codex, copilot: fully synced").
Refactored show*Diff → collect*Diff to separate data collection
from rendering, enabling cross-target fingerprint grouping via
renderGroupedDiffs().
Keyboard always controls left tree (j/k navigate with auto-preview,
h/l collapse/expand). Ctrl+d/u/g/G always scroll right content panel.
No focus switching needed — eliminates sidebarFocused/viewingFile state.
Dual-pane content viewer for list TUI: left sidebar shows recursive file
tree (depth≤3), right panel renders Markdown via glamour with custom style.
Vim-like navigation: h/l expand/collapse dirs, j/k auto-preview, Tab to
switch panels. Mouse wheel scroll and click supported.
- Report "partial" status for sync/install/update/uninstall when some
items succeed and others fail
- Instrument 6 new commands: check, diff, trash, init, upgrade, and
all mutating commands now log to operations.log
- Auto-truncation via WriteWithLimit with configurable log.max_entries
(default 1000) and 20% hysteresis to avoid truncating every write
- Stats engine: computeLogStats with per-command breakdown
- CLI --stats flag with colored output and ✓ok/total ratios
- TUI stats panel (toggle with 's'): bar charts, status distribution,
last operation timestamp
- TUI stats footer: compact always-visible summary line
- TUI detail panel: j/k scrolling for long entries, proportional
list/detail layout (40%/60%)
- Web API: GET /api/log/stats endpoint
- Updated website docs for log command
- Add CHANGELOG entry for v0.16.2 (Azure DevOps, --prune, stale detection, Windows ARM64)
- Update install.md with Azure DevOps source format section, Azure Pipelines CI example
- Add Azure DevOps to intro.md "Any Git Host" feature bullet
- Trim SKILL.md description to 218 chars, bump version to v0.16.2
Add PlatformAzureDevOps detection (dev.azure.com, ssh.dev.azure.com,
visualstudio.com) and AZURE_DEVOPS_TOKEN env var support. Token is
auto-injected via git config insteadOf rewrite, same pattern as
GITHUB_TOKEN/GITLAB_TOKEN/BITBUCKET_TOKEN.
Update error messages, sanitizeTokens redaction, environment-variables
docs with PAT creation guide and official Microsoft docs link, and
url-formats docs with Azure DevOps auth section.
Azure DevOps uses 4-segment paths (org/project/_git/repo) which the
existing 3-segment gitHTTPSPattern misparses. Add dedicated patterns
for modern HTTPS (dev.azure.com), legacy HTTPS (visualstudio.com),
and SSH v3 (ssh.dev.azure.com) formats. Also add ado: shorthand
(e.g. ado:org/project/repo) for quick installs.
Includes TrackName support (org-proj-repo format), 16 new test cases,
and url-formats.md documentation with platform support table update.
- Extend .skillshare-manifest.json tracking from copy mode to merge mode.
- Update PruneOrphanLinks to prioritize manifest records over naming patterns.
- Ensure SyncTargetMerge writes to manifest for all linked/updated skills.
- Add integration tests for uninstall + sync orphan directory cleanup.
- Maintain backward compatibility for targets without existing manifests.
- CLI auth hint now shows PowerShell syntax on Windows, export on Unix
- Docs: add Windows PowerShell examples for GITLAB_TOKEN, BITBUCKET_TOKEN,
SKILLSHARE_GIT_TOKEN in environment-variables.md
- Docs: rewrite auth error section in common-errors.md with token env var
options, Windows syntax, and GitLab token type warning
Automatically detect GITHUB_TOKEN, GITLAB_TOKEN, BITBUCKET_TOKEN, or
SKILLSHARE_GIT_TOKEN and inject credentials via transient GIT_CONFIG_*
env vars during clone and pull. Tokens are never persisted to disk.
Key changes:
- Platform detection (GitHub/GitLab/Bitbucket) with per-platform
username conventions (x-access-token, oauth2, x-token-auth)
- GIT_CONFIG_COUNT safe append — reads existing count to avoid
overriding CI pipeline git config entries
- Token-aware error messages: "no token" shows 3 options (SSH, env
var, credential helper); "token rejected" suggests checking
permissions and expiry
- sanitizeTokens redacts all credential values (including
BITBUCKET_USERNAME) from stderr before surfacing errors
- gitPull also injects auth via getRemoteURL(), fixing the previous
limitation where update failed after HTTPS+token install
Enable IDE autocompletion, validation, and hover docs for config.yaml
by adding JSON Schema support via yaml-language-server directive.
- Add schemas/config.schema.json (global) and project-config.schema.json
- Prepend schema comment in Config.Save() and ProjectConfig.Save()
- Add unit and integration tests for schema comment roundtrip
- Document IDE setup in configuration.md
Change matchSkillIgnore to compare against skill path instead of name,
enabling patterns like "feature-radar" to exclude all skills under
that group directory. Precise paths like "feature-radar/helper" still
target individual skills. Backward compatible — flat skills have
path == name so existing patterns work unchanged.
Add step-by-step false positive fix workflow with scaffold file
preview to audit docs. Add directory grouping section to list docs
explaining auto-grouping mechanism and linking to organizing guide.