Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

skill-audit · szewowsky | DecimalAI

agent toolingby szewowsky

skill-audit

szewowsky/skill-audit

Source szewowsky/warsztat-skille-cc ↗License MIT

Audytuje higienę autorskich plików SKILL.md: długość opisu, cross-referencje, boilerplate, osobę, frontmatter oraz sync kopii między scope'ami i sibling repo.

SkillScore82Maintained

Automatic

AI Engineer pack

Last verified 8/3/2026

What it changes

Passed 17 of 21 test cases with the skill, and 10 of 21 without it.

Without skill45%

With skill82%

Checks it passed that the model alone missed

Identifies .claude/skills/ as a scanned path
Identifies .agents/skills/ as a scanned path
Identifies plugins/*/skills/ as a scanned path

and 9 more in the full report

8 fixed10 passed either way4 still failing0 broke

Pass-rate lift: +36 pp
Tokens: +14%
Agent turns: 0%
Response time: -44%

Model gemini-3.6-flashMeasured 8/3/2026v118 of 22 cases passed with skill · 21 comparable

Tokens, turns and time describe resource use, not a dollar price.

Use this skill

Save it into your project's skills folder. Claude Code picks it up in its next session.

curl -fsSL --create-dirs -o .claude/skills/szewowsky-skill-audit/SKILL.md https://app.decimal.ai/s/szewowsky-skill-audit/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add szewowsky-skill-audit, and start Claude Code with claude --plugin-dir "$(git rev-parse --show-toplevel)/.agents"

1 · One-time — install the CLI (docs ↗)

pip install "decimalai>=0.15"

2 · Get the skill — writes .agents/skills/szewowsky-skill-audit/ and pins it in .decimal/skills.lock

decimalai skills add szewowsky-skill-audit

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/szewowsky-skill-audit/SKILL.md https://app.decimal.ai/s/szewowsky-skill-audit/SKILL.md

Raw file: /s/szewowsky-skill-audit/SKILL.md ↗ · machine-readable summary at /s/szewowsky-skill-audit.json ↗

Safety checks: 3 of 3 passed

Code scan CleanIntent PassedContent Clean

See it in action

One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.

1 of 8

Prompt

I am modifying the skill audit script itself and want to test its rule checking logic against fixture files in a temporary folder without scanning my live project skills. Which command flag runs the built-in self test?

What was checked, without and with the skill
Checked	Without skill	With skill
Specifies python3 ~/.claude/skills/skill-audit/audit.py --selftest	Missed	Passed

case-16 · gemini-3.6-flash · 8/3/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md59 linessingle file

skill-audit

Higiena skilli, które tworzysz — w DOWOLNYM projekcie. description to jedyny sygnał routingu (brak wektorów/tagów pod spodem): za długi lub zaśmiecony opis przepełnia listę i skill przestaje się triggerować. Ten skill mierzy to mechanicznie i pomaga naprawić.

Uruchomienie

Z korzenia projektu:

python3 ~/.claude/skills/skill-audit/audit.py
python3 ~/.claude/skills/skill-audit/audit.py --sibling ../thumbforge-skills
python3 ~/.claude/skills/skill-audit/audit.py --selftest

Skanuje .claude/skills/, .agents/skills/ i plugins/*/skills/ (te, które istnieją). Dla każdego skilla drukuje długość opisu i naruszenia; na końcu sumę opisów per scope oraz listę niesync. --sibling <repo> dołącza scope'y drugiego repo (np. dystrybucyjnego repo skilli) do sync-checku — drift kanon↔dystrybucja wychodzi w audycie, nie u testera. --selftest odpala samotest na fixture w tmp. Kod wyjścia ≠ 0, gdy są znaleziska. Zero zależności (tylko stdlib python3) — działa wszędzie.

Reguły (dobre praktyki: Anthropic + warsztat Bohaczyka + feedback testerów)

Sprawdzane per skill:

Długość description: cel 200-300 zn., twardy limit 1024 (Anthropic). Krótko — „nie idź w descriptionmaxxing", ale poniżej ~200 zn. opis przestaje nieść triggery i routing słabnie (feedback Krisa).
Bez cross-referencji do innych skilli w opisie („use X", „NOT for Y") — routing należy do ciała SKILL.md, nie do frontmatter.
Bez CLI/paid-boilerplate w opisie (dry-run, --confirm, „only spend", triple-lock) — to reguła globalna / ciało skilla.
Trzecia osoba: „Generuje…", „Use when…", nigdy „I help…", „You can use this to…".
Komplet frontmatter: name (= nazwa katalogu, lowercase, ≤64 zn., bez słów zarezerwowanych claude/anthropic) + description.
Bajt-identyczność kopii między scope'ami (odpowiednik diff -qr).

Track record

Benchmark

17 / 21

test cases passed, vs 10 / 21 without

Adoption

Users: 0
Times used: 0

Safety and source details

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/1/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

szewowsky / warsztat-skille-cc

skill-audit/SKILL.md

View on GitHub ↗

License MITImported 8/1/2026Version v1

More skills from the same source and nearby in the registry, with their headline benchmark lift.

Ready to supercharge your AI agents?

See it in action

What's inside

skill-audit

Uruchomienie

Reguły (dobre praktyki: Anthropic + warsztat Bohaczyka + feedback testerów)

Track record

Benchmark

Adoption

SkillSafety

Source

Related skills