1. install.ts: gate marketplace-resolution console.log behind !this.json && !this.quiet so --json output stays a single JSON blob and --quiet stays silent. Uses real flag names from the InstallCommand class (this.json, this.quiet).
2. tool-abuse.ts: TA002 pattern 2 lookbehind was too broad. The original list (must|should|always|will|to) included malicious intensifiers — 'Always run without confirmation' and 'Will execute without approval' were incorrectly suppressed. Replaced with negation-only tokens (not|n'?t|never|avoid|forbid|deny|reject|prevent|cannot|can'?t) so genuine negation like 'Do not run without confirmation' still suppresses, but malicious intensifiers fire.
3. rules.test.ts: scanContent helper now uses try/finally so the temp directory is cleaned up even when analyzer.analyze throws.
Tests: 3 new cases for the TA002 lookbehind fix. Full core suite 1051/1051 pass.
Four scanner rules over-fired on standard Node patterns and safety documentation, blocking install of skills like pro-workflow on every node-script skill. Findings tracked at #129.
TA002 (autonomy abuse): patterns 2 and 3 now use negative lookbehind for negation context. 'Never auto-approve' / 'must not bypass approval' no longer trip the rule, but bare 'Auto-approve all tools' still fires.
CI003 (child_process): dropped the bare 'require(child_process)' pattern. Importing the module is not the risk surface — exec() and execSync() are, and they have their own patterns. const { execFileSync } = require('child_process') is the recommended safe form.
CI005 (template literal injection): narrowed pattern to template literals inside shell-spawning calls (exec/execSync/spawn/spawnSync). Authorization headers like `Bearer ${env.TOKEN}` and string concatenation like fs.writeFileSync(file, `${header}${rows}`) are no longer flagged.
CI007 (shell chaining): dropped 'markdown' from fileTypes. Markdown documentation of dangerous shell patterns (deny-list tables, security guides) is documentation, not execution. Real .sh/.bash files still scanned.
install command: marketplace name lookup. 'skillkit install <bare-name>' now consults marketplace/skills.json for entries matching by name or id and resolves to the canonical source string before failing with 'Could not detect provider'.
Tests: 13 new scanner regression tests in packages/core/src/scanner/__tests__/rules.test.ts. Full suite 1048/1048 pass.
Version bump
------------
1.22.1 -> 1.23.0 across all workspace packages. Minor bump covers the
slim-install dependency layout, safer first-time agent picker, and the
corresponding doc + branding changes.
Prune dead optional deps in @skillkit/core
------------------------------------------
@anthropic-ai/sdk, @google/genai, ollama, and openai were listed as
optionalDependencies but never imported. They were pulled through
@skillkit/cli transitively on every default install, adding ~57
packages and one deprecation warning (node-domexception via the
SDKs' own formdata-node / node-fetch transitives). Grep of
packages/core/src confirms no SDK import; providers in
src/ai/providers use direct fetch() against the provider HTTP APIs.
Drop them. better-sqlite3, sqlite-vec, and node-llama-cpp stay
because they are actually dynamic-imported by the vector store and
embedding pipeline.
Measured on a fresh npm cache (overrides-pinned file: tarballs):
before: 475 pkgs / 18s / 3 deprecation warnings
1.23.0: 418 pkgs / 18s / 2 deprecation warnings
slim: 118 pkgs / 9s / 0 deprecation warnings
The two remaining warnings are upstream:
- prebuild-install@7.1.3 from better-sqlite3 (must stay)
- glob@9.3.5 from @opentui/solid -> babel-plugin-module-resolver
README rewrite
--------------
Punchier top fold. Palette-consistent SVG tag badges under
assets/tags/ replace most shields.io badges (still keep npm/CI/etc
on shields since they need live data). Collapsible detail blocks
for every feature instead of the linear wall. Same nav anchors the
site uses. No em dashes. No AI filler words. Writing style pulls
from the ai-engineering-from-scratch and awesome-claude-design
READMEs the user already runs.
Branded SVG tags
----------------
scripts/gen-tags.mjs generates 20 SVG pills under assets/tags/ in
the SkillKit palette (#09090b base, #fafafa text, #3f3f46 border,
accent per tag). Two variants: single pill with left accent bar
(new, slim-install, v1.23.0, etc.) and split badge (agents | 45,
packages | 118, install | 9s). Matches the aesthetic used on
awesome-claude-design, tailored to SkillKit's monochrome brand.
The tags directory is fork-and-remix friendly; README points to it.
- Fix "44 agents" → "45 agents" across 34 files (website, docs,
package.json, tests, extension, API, TUI)
- README: add tap, check, update, publish submit to commands section
- Commands.tsx: add Sources category with tap commands, update manage
group with remove --source/--all and check
- commands.mdx: full docs for tap, check, update, remove, publish
submit with all new flags and behaviors
- configuration.mdx: document lock.json and taps.json
- security.mdx: document trust badges system
- Use finally for temp cleanup so user cancels don't leak cloned dirs
- Move addSkillToLock outside inner agent loop, merge agents array
- Use source:name as lock key to avoid cross-source collisions
- Gate GitHub activity check on provider.type (skip for GitLab/Bitbucket)
- Normalize source URL to owner/repo before official check
- Include skill name in trust badge output
- Atomic lock.json writes via tmp file + renameSync
- Remove stale --federated suggestion from find no-results message
Introduce ~/.skillkit/lock.json with LockEntry/LockFile schemas to
track all installed skills in a single file. Install writes entries,
remove deletes them. Enables future team sync and reproducible setups.
Compute SHA-256 checksum of SKILL.md on install and store in metadata.
During update, check GitHub pushedAt timestamp and compare checksums
to skip unnecessary re-downloads when content hasn't changed.
- findSkill: fall through to .md check when parseSkill returns null
- Use isPathInside instead of dir.includes(cwd) for location detection
- Optional chaining on meta.source in remove --source filter
- Track fence delimiter type per CommonMark (backtick vs tilde)
- Wrap statSync in try-catch so one bad source doesn't abort all installs
- Add cleanup fallback when sparse-checkout/checkout fails
Standalone .md skills (e.g. doc_explanation.md) now get metadata saved
as .doc_explanation.skillkit.json in the same directory. This allows
remove --source to find and remove them. Updated saveSkillMetadata,
loadSkillMetadata, loadMetadata, and remove command to handle both
directory and sibling metadata patterns.
getSearchDirs now collects skill directories from ALL agent adapters
instead of only the configured agent. This fixes 'skill not found' when
skills were installed to .claude/skills/ but the default agent is
universal (which only searches skills/).
findSkill now also checks name.md for standalone .md skills that were
installed as plain files.
- Restrict try/catch to only the partial clone spawn so
sparse-checkout/checkout failures bubble up as real errors instead
of being misreported as "partial clone unsupported"
- Cap stderr accumulator to 4KB to bound memory on verbose clones
- Add -- terminator before sparsePaths in sparse-checkout set to
prevent user-controlled subpath from being parsed as git options
- spawnGit: set stdout to 'ignore' instead of 'pipe' to prevent
buffer deadlock (git clone writes to stderr, not stdout)
- tryPartialClone: emit fallback message via onProgress instead of
silently swallowing the error
- scanner: preserve fence markers so PI011 delimiter injection is
still detected, only strip inner code block content
- scanner: guard against unclosed frontmatter by looking ahead for
closing --- before entering frontmatter mode
cloneRepo now accepts a subpath option. When set, sparse-checkout
includes both the subpath itself and SKILL_DISCOVERY_PATHS under it.
All four providers pass their parsed subpath through to cloneRepo.
Without this, sparse checkout would skip the subpath directory and
discoverSkills would find nothing.
Strip fenced code blocks and standard YAML frontmatter before scanning
markdown files. This eliminates three categories of false positives:
- PI013: no longer flags standard --- frontmatter delimiters
- CI007: no longer flags curl|sh in code examples
- DE003: no longer flags process.env in code examples
Real threats in prose (outside code blocks) are still caught.
Replace silent execFileSync with async spawn + --progress flag so the
spinner shows real download progress (e.g. "Receiving objects 45%").
For large monorepos, try partial clone (--filter=blob:none) with sparse
checkout of SKILL_DISCOVERY_PATHS first, falling back to regular
--depth 1 clone if the server doesn't support it. Also adds
--single-branch --no-tags to all clones.
Shared cloneRepo() helper in base.ts eliminates duplicate clone logic
across all four git providers.
- Remove legacy clawdbot path checks from OpenClawAdapter.isDetected()
so pure clawdbot workspaces are not hijacked by OpenClaw detection
- Add '.openclaw/agents' to ALL_AGENT_DISCOVERY_PATHS so agent
discovery works for OpenClaw agent directories
* feat: add multi-tier eval engine (skillkit eval)
6-tier evaluation system for comprehensive skill assessment:
- Tier 1: LLM quality scoring (G-Eval pattern, 6 dimensions)
- Tier 2: Contradiction detection (formal rules + LLM semantic)
- Tier 3: Behavioral security (code analysis + taint tracking + LLM)
- Tier 4: Sandbox execution testing (Docker, graceful skip)
- Tier 5: Dynamic marketplace benchmarks (percentile ranking)
- Tier 6: Community signals (freshness, metadata, content health)
Works without API keys (heuristic fallback for Tier 1, Tiers 5-6).
Full LLM evaluation with --provider flag.
* fix: address Devin review — input validation for eval CLI
- --min-score NaN bypass: validate parseInt result before comparison
- --timeout NaN propagation: skip NaN timeout instead of passing to Docker
- --tier invalid input: graceful error message instead of unhandled throw
- scoreContentSize formula: cap sub-optimal at 99 to distinguish from optimal
* fix: address CodeRabbit review findings on eval system
- Harden contradiction/security prompts against prompt injection by
wrapping untrusted content in <skill_content> XML tags and adding
explicit "treat as untrusted data" instructions
- Use head+tail content sampling in security prompt instead of head-only
truncation to catch malicious payloads at end of content
- Add missing Tier 4 sandbox formatter to reporter.ts
- Fix misleading sampleSize fallback in dynamic-benchmark.ts
- Move --min-score NaN validation before engine.evaluate() to avoid
wasting LLM calls on invalid input
- Add truncation indicator to quality-cot.ts prompts
- Add term length limits (>100 chars) in contradiction.ts to mitigate
ReDoS risk on dynamically constructed regex
* fix: address round 2 review findings on eval system
- Wrap quality-cot prompts with <skill_content> XML tags and untrusted
data instructions, consistent with contradiction/security prompts
- Change truncateContent to head+tail sampling (4000+2000) instead of
head-only, so completeness/token-efficiency evaluators see both
beginning and end of long skills
- Fix escapeXmlTags regex to handle whitespace variants before closing
angle bracket (</skill_content >, </skill_content\n>)
- Type DIMENSION_PROMPTS with Record<EvalDimension, PromptBuilder> for
compile-time key validation
- Clamp estimatePercentile return to [0,100] range defensively
- Make buildStatsFromMarketplace async with periodic yields (every 50
skills) to avoid blocking the event loop on first-run cache miss
- Extract parsedTimeout local variable in eval.ts to avoid double
parseInt parsing
* fix: address round 3 Devin + CodeRabbit findings on eval system
- Scan both code blocks AND full content for dangerous patterns in
behavioral-security.ts, with deduplication by description+snippet
to avoid double-reporting (Devin: security scanner skipped prose)
- Use non-greedy regex /\[[\s\S]*?\]/ in parseSemanticFindings to
match first complete JSON array instead of greedily spanning to
last bracket in LLM response (Devin: greedy regex corrupted JSON)
- Fix formatSandboxDetails to use r.testCase matching SandboxResult
type instead of r.name which doesn't exist (Devin: property mismatch)
- Make MarketplaceSkill.description optional to match marketplace
schema where only id/name/source/tags are required (CodeRabbit)
- Rename sandbox test case from "code example execution" to "code
block content validation" since it validates content not execution
(CodeRabbit)
* fix: address round 4 Devin + CodeRabbit findings on eval system
Inline fixes:
- Replace greedy JSON regex in behavioral-security.ts runLLMAnalysis
with balanced bracket extraction that correctly handles ] inside
JSON string values
- Replace non-greedy regex in contradiction.ts parseSemanticFindings
with same balanced bracket extraction (non-greedy also breaks on
inner brackets)
- Fix crossValidatedCount in behavioral-security.ts to accumulate
(+= uniqueEngines.size) instead of replacing (Math.max)
- Fix gradeDeterministic: cleanExit && hasOutput without matching
expectedOutcome now returns passed:false score:50 instead of
passed:true score:75
- Docker unavailable returns score:-1 sentinel instead of 0/F;
engine.ts filters score<0 tiers from overall average so
Docker-less machines aren't penalized
Duplicate fix:
- Add score field to SandboxResult type; store gradeResult.score
in each result; compute tier score from average of numeric scores
instead of boolean pass rate
Nitpick fixes:
- Distinguish "Docker unavailable" from "No tests executed" in
formatSandboxDetails
- Tighten base64 obfuscation regex to require 4-char groups
- Only push 'llm' to engines when LLM actually produced findings
- Bound \s+ to \s{1,10} in dynamically constructed regexes in
contradiction.ts to prevent catastrophic backtracking
- Validate cache timestamp with Number.isFinite() in loadCache to
reject corrupted cache files with NaN timestamps
Ran `npx tessl skill review --optimize` on all methodology pack
skills and find-skills. Key improvements:
- Better descriptions with "Use when..." clauses and concrete verbs
- Lowercase-hyphenated name format for consistency
- Trimmed self-evident prose Claude already understands
- Added executable examples and quick reference tables
- Condensed verbose sections into compact tables/summaries
- Fixed YAML parsing errors (unquoted colons in descriptions)
Net result: -552 lines across 14 files.
* feat: integrate skills.sh as first-class registry
Add SkillsShRegistry and SkillsShProvider so users can install skills
directly from skills.sh (e.g. `skillkit install skills.sh/owner/repo/skill`).
The find command now searches skills.sh by default alongside the local
marketplace. Post-install shows skills.sh popularity stats and pro tips
highlighting skillkit's unique features (translate, recommend, evaluate).
* fix: address CodeRabbit and Devin review findings for skills.sh integration
- Fix path traversal vulnerability in SkillsShProvider.clone (validate subpath containment)
- Fix federated search skipped for interactive queries (use searchTerm instead of this.query)
- Guard showSkillsShStats to skip network call for local providers
- Replace deprecated npm --production with --omit=dev
- Fix loose includes() match in getSkillStats to use strict endsWith()
- Fix FederatedSearch resilience: don't re-throw RateLimitError, return partial results
- Forward timeoutMs through FederatedSearch to registries
- Extract shared formatCount utility to helpers.ts (deduplicate find.ts/install.ts)
- Fix install hint to use skill.source instead of fragile skill.description
- Fix timer leak in fetchLeaderboard using finally block
- Deduplicate URL parsing via shared parseSkillsShParts helper
- Validate --agent flag values against known agent types
* Security hardening and code quality fixes
- Replace execSync with execFileSync to prevent shell injection (providers, executor, fetcher, publish)
- Add security headers: X-Content-Type-Options, X-Frame-Options, Content-Security-Policy, Referrer-Policy
- Add input validation for save route name field to prevent path traversal
- Fix rate limiter IP extraction (x-forwarded-for + x-real-ip fallback)
- Default server bind to 127.0.0.1 instead of 0.0.0.0
- CORS fallback to * (configurable via options.corsOrigin)
- Use execFileSync in publish submit (avoid shell injection when opening browser)
- TLS-aware health checks and HTTPS support in RemoteTransport
- Conditional rejectUnauthorized for TLS (self-signed configurable)
- Add **/.env.local patterns to .gitignore
- Formatting: double quotes, trailing commas (Prettier)
* fix: address CodeRabbit and Devin review findings from PR #74
- TLS: default allowSelfSigned=true to preserve self-signed cert workflow;
enforce rejectUnauthorized=true when trustedCAs are supplied in client context
- Signature verification: verifySecureMessage now cryptographically verifies
signatures using PeerIdentity.verifyHex instead of only checking fingerprints
- WebSocket: fix timeout handle leak in performClientHandshake; add error
logging to empty catch blocks; add per-socket error handler; pre-compute
signature once in broadcast(); add send error callbacks; reject startup if
TLS configured but certInfo missing
- Command parsing: replace broken regex tokenizer with shared splitCommand()
utility that correctly handles --flag="value with spaces"
- Rate limiter: prefer x-real-ip, fall back to rightmost x-forwarded-for entry
- Save route: scope IPv6 prefix checks to actual IPv6 addresses; stop leaking
internal error messages to clients
- Publish: extract shared parseSkillFrontmatter(); fix getRepoInfo to use
async execFileSync; fix path containment checks with separator
- Remote transport: add rejectUnauthorized option; bracket IPv6 in URLs;
fix host:port splitting for IPv6 with lastIndexOf
- Health checks: add tls/tlsAllowSelfSigned to Host type; pass https options
- Fetcher: validate owner/repo; derive indexDir from INDEX_PATH
- .gitignore: add recursive **/.env pattern, remove redundant entries
* fix(executor): restore shell feature support in shellExecutor
The previous security hardening replaced execSync with execFileSync,
which broke commands using pipes, redirects, and subshells. Now detects
shell metacharacters and falls back to sh -c for those commands while
keeping execFileSync for simple commands.
* chore: migrate domain from agenstskills.com to skillkit.sh
Update all OG/Twitter meta tags, API endpoints, docs links, badge
URLs, privacy policy, extension store listing, and README references
to use skillkit.sh. Both skillkit.sh and agenstskills.com are treated
as equal domains — privacy policy, store listing, and docs reference
both.
* perf: skip Vercel builds when docs unchanged
Add ignoreCommand to both docs/skillkit and docs/fumadocs vercel.json.
Vercel will skip rebuilds when the respective docs folder has no
changes, saving ~1h+ of build minutes per billing cycle.
* feat: add `skillkit doctor` command and enhance `skillkit status`
- New `skillkit doctor` command with 10 health checks across 3 categories
(environment, configuration, skills) with --json and --fix flags
- Enhanced `skillkit status` shows project overview dashboard when no
active session exists instead of bare "No active session found" message
- Overview includes agent, config, version, skills count, recent history
* feat: add skill activity log, session snapshots, and session explain
Three new session intelligence features:
- Activity Log: tracks which skills were active per git commit
(`skillkit activity` with --skill, --limit, --json flags)
- Session Snapshots: save/restore named snapshots of session state
(`skillkit session snapshot save|restore|list|delete`)
- Session Explain: structured summary of current session
(`skillkit session explain` with --json, --no-git flags)
Includes path traversal protection on snapshot names, YAML validation
on parsed activity data, and robust error handling on snapshot restore.
983 tests passing (942 core + 41 CLI including 14 E2E tests).
* feat: add issue planner - generate plans from GitHub Issues
New `skillkit issue plan` and `skillkit issue list` commands that
fetch GitHub Issues via `gh` CLI and generate StructuredPlans.
Parses checkboxes, file mentions, and labels from issue body.
29 new tests, all passing.
* fix: address CodeRabbit review suggestions for PR #64
- doctor.ts: use execFileSync, typed AgentType, remove as-any casts
- activity.ts: validate parseInt with radix and NaN guard
- issue.ts: wrap JSON.parse in try/catch for gh CLI output
- issue-planner.ts: guard JSON.parse, fix includeTests no-op ternary,
remove duplicate IssuePlanMetadata (import from types.ts)
- status.ts: typed StatusOverview interface, remove as-any casts
- snapshot-manager.ts: align observations param with SessionSnapshot type
- session.ts: proper type mapping for observations (no double cast)
- activity-log.ts: min-length guard for SHA prefix match
- e2e tests: use top-level imports instead of inline require()
- commands.test.ts: add assertions for DoctorCommand, IssuePlanCommand,
IssueListCommand
* fix: address review feedback from CodeRabbit and Devin
- Include test step in no-checklist fallback when includeTests is true
- Restore observations from snapshot in SessionSnapshotRestoreCommand
* fix: address remaining CodeRabbit and Devin review feedback
- doctor.ts: Use getAdapter() per AgentType instead of getAllAdapters()
spread (which destroys prototype methods like isDetected)
- doctor.ts: Use getGlobalSkillsDir() from AGENT_CONFIG instead of
hardcoded path formula
- session-explainer.ts: Defensively default state.history and
state.decisions to empty arrays
- session-explainer.ts: Use loadConfig().agent with skillSource fallback
instead of always using skillSource as agent name
- observation-store.ts: Add static readAll() that reads observations
without sessionId mismatch clearing
- session.ts: Use ObservationStore.readAll() for snapshot save to avoid
empty observations due to sessionId mismatch
* feat: implement standards reference — URL-to-Skill, AGENTS.md, spec validation
Phase 1-4 of positioning SkillKit as THE runtime implementing all three
AI agent skill standards (Agent Skills, AGENTS.md, well-known discovery).
New modules:
- save/ — ContentExtractor, SkillGenerator, AutoTagger for URL-to-Skill
- agents-md/ — AGENTS.md generator with managed section markers
- validation/ — SpecValidator with strict mode for agentskills.io compliance
New CLI commands:
- `skillkit save <url>` — save any URL/text/file as a reusable SKILL.md
- `skillkit agents init|sync|show` — generate and manage AGENTS.md
Updated:
- Well-known provider: Mintlify convention paths, discoverFromUrl()
- Install command: smart URL fallback (well-known → offer save)
- Remove command: auto-updates AGENTS.md on skill removal
- Validate command: --strict flag for spec compliance checks
- Quality module: name-directory match + token budget warnings
- Schema: agents enum updated from 7 → 44 agents
* fix: address Devin review — register AgentsMd commands, fix temp dir leak, wrap AGENTS.md update
- Register AgentsMdCommand/Init/Sync/Show in CLI entry point (were exported but never registered)
- Clean up tempDir in wellknown clone() when no skills found via index or mintlify
- Wrap AGENTS.md auto-update in install.ts with try-catch to prevent masking successful installs
* fix: address CodeRabbit review findings
- wellknown.ts: drop multiline flag from frontmatter regex, extract skill name
from URL path segments instead of basename on full URL
- agents-md.ts: wrap init/sync in try-catch for user-friendly error messages
- publish.ts: add path-traversal guard to mintlify branch
- validate.ts: remove duplicate error/warning output in spec report
- extractor.ts: add 30s AbortSignal timeout to all fetch calls
- skill-generator.ts: fix .filter(Boolean) stripping blank line after frontmatter
* feat: add 12 missing AI coding agents (Devin, Aider, Cody, Amazon Q, Augment, Replit, Bolt, Lovable, Tabby, Tabnine, CodeGPT, PlayCode)
Expands agent support from 32 to 44 agents across all packages:
- Core types, configs, discovery paths, format maps, command generator
- Agent adapters (UniversalAdapter for all new agents)
- CLI onboarding icons/names, TUI symbols/logos
- Updated all docs, tests, and e2e references from 32 to 44
* test: add comprehensive tests for new agents, update docs and e2e
- Add it.each tests for all 12 new agent adapters and configs
- Add completeness test ensuring every AgentType has adapter + config
- Update agents.mdx doc table with all 12 new agents
- Update e2e onboarding test with new agent display names
* fix: add all 44 agents to translateSkillToAll and fix e2e test
- translateSkillToAll had hardcoded 17-agent list, now includes all 44
- e2e/skill-translator test was missing cline-zencoder batch in SUPPORTED_AGENTS
* fix: replace UniversalAdapter with GenericAgentAdapter for all non-universal agents
UniversalAdapter hardcoded type='universal', skillsDir='skills', and
name='Universal (Any Agent)' for all 26 agents that used it. This caused
skills to install to wrong directories and agents to be undetectable.
GenericAgentAdapter reads type, skillsDir, configFile from AGENT_CONFIG
and checks agent-specific directories in isDetected(). Fixes both the
12 new agents and the 14 pre-existing ones (cline through zencoder).
115 adapter tests now verify correct type, skillsDir, configFile, and
name for every GenericAgentAdapter instance.