{ "markdown": "# pulser — SKILL.md Linter for Claude Code\n\n

\n \"pulser\n

\n\n

\n Website ·\n npm ·\n 한국어\n

\n \"CI\"\n \"npm\"\n \"downloads\"\n \"license\"\n \"stars\"\n \"Mentioned\n \"GitHub\n

\n\n\"Check my skills.\" Diagnose, classify, prescribe, and fix — without leaving the conversation.\n\n```\n$ pulser\n\n pulser v0.4.0\n\n 54 skills scanned · Score: 89/100\n ✓ 48 healthy ⚠ 4 warnings ✗ 2 errors\n\n Top issues:\n cardnews — No Gotchas, no allowed-tools\n geo-audit — 338 lines, single file\n\n 💊 Rx #1 — cardnews\n [GOTCHAS] Add Gotchas section\n Why: Anthropic's highest-ROI improvement\n Template:\n ## Gotchas\n 1. Validate output against conventions\n 2. Check scope — don't over-generate\n\n Fix type: AUTO\n```\n\n## What it does\n\npulser scans your `SKILL.md` files against 8 diagnostic rules derived from Anthropic's published principles in [\"Building Claude Code: How We Use Skills\"](https://code.claude.com/docs/en/skills):\n\n| Rule | What it checks |\n|------|---------------|\n| `frontmatter` | Required `name` and `description` fields |\n| `description` | Trigger keywords, \"Use when\" pattern, length |\n| `file-size` | SKILL.md under 500 lines |\n| `gotchas` | Gotchas section with failure patterns |\n| `allowed-tools` | Tool restrictions appropriate for skill type |\n| `structure` | Supporting files for large skills |\n| `conflicts` | Trigger keyword overlap between skills |\n| `usage-hooks` | Skill usage logging hook installed |\n\nEach skill is auto-classified by type (analysis, research, generation, execution, reference) with confidence scoring. Prescriptions are tailored to the detected type.\n\n## Install\n\n```bash\nnpm install -g pulser-cli\n```\n\nOn install, pulser registers itself as a Claude Code skill — say \"check my skills\" or `/pulser` to run it conversationally.\n\n## Usage\n\n### In Claude Code\n\nJust say it:\n\n```\ncheck my skills\n```\n\nOr use the slash command:\n\n```\n/pulser\n```\n\nClaude runs the diagnosis, summarizes results, and offers to fix issues — all within the conversation.\n\n### In terminal\n\n```bash\n# Scan default path (~/.claude/skills/)\npulser\n\n# Scan a specific directory\npulser ./my-skills/\n\n# Scan a single skill\npulser --skill reasoning-tracer\n\n# Auto-fix issues with backup\npulser --fix\n\n# Rollback the last fix\npulser undo\n\n# JSON output (for CI/automation)\npulser --format json\n\n# Markdown report\npulser --format md\n\n# Treat warnings as errors\npulser --strict\n\n# Disable TUI animation\npulser --no-anim\n```\n\n## Eval — Skill Testing\n\nNew in v0.4.0: test your skills against real inputs.\n\n```bash\npulser eval\n```\n\nWrite `eval.yaml` next to your `SKILL.md`:\n\n```yaml\ntests:\n - name: \"catches bugs\"\n input: \"Review: function add(a,b) { return a - b }\"\n assert:\n - starts-with: \"The bug\"\n - contains: \"subtract\"\n - min-length: 30\n```\n\npulser runs each test through `claude -p`, checks assertions, and tracks regressions automatically.\n\n```\n$ pulser eval\n\n reviewer (2 tests)\n ✓ catches bugs 320ms\n ✓ passes clean code 280ms\n\n 2 passed · 0 failed · 0.6s\n```\n\nSupported assertions: `contains`, `not-contains`, `starts-with`, `ends-with`, `min-length`, `max-length`, `matches` (regex).\n\n### Exit Codes (eval)\n\n| Code | Meaning |\n|------|---------|\n| `0` | All tests passed |\n| `1` | Failures found |\n| `3` | Regression detected (previously passing test now fails) |\n\n## Core Pipeline\n\n1. Diagnose — Scan and classify issues across 8 rules\n2. Prescribe — Explain why it matters, provide ready-to-use templates\n3. Fix — Auto-apply safe structural fixes with full backup\n4. Eval — Test skills against real inputs, track regressions\n5. Rollback — Instant undo, your safety net\n\n## Exit Codes\n\n| Code | Meaning |\n|------|---------|\n| `0` | All rules passed |\n| `1` | Errors found |\n| `2` | Warnings found (with `--strict`) |\n\n## Patient Monitor TUI\n\nWhen running in a TTY terminal, pulser displays a hospital-style patient monitor with real-time waveform animation:\n\n- Green ECG — Skills being scanned\n- Green capnography — Pass/warn/fail counts\n- Cyan plethysmograph — Health score\n- Yellow respiratory — Prescription count\n\nDisable with `--no-anim` or pipe to a file.\n\n## License\n\nMIT — [TheStack-ai](https://github.com/TheStack-ai)\n\n## Featured in\n\n- [GitHub Marketplace — Pulser Action](https://github.com/marketplace/actions/pulser-claude-code-skill-linter)\n- [Awesome Claude Code Toolkit](https://github.com/rohitg00/awesome-claude-code-toolkit)\n- [Anthropic Plugin Directory](https://platform.claude.com) *(under review)*\n", "bytes": 5777, "sha": "505d18520ab0017f48dc9ee209da812260ff59e5b5832da0304ae3d975635bea", "repo_slug": "thestack-ai/pulser", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/plg_thestack_ai_pulser_pulser_adfea54e/readme" }