I installed a bunch of agent skills. Then I wrote a tool to audit them.

TECHDEV — Sat, 17 Oct 2026 10:02:54 +0000

Every week a new agent skills pack shows up on GitHub Trending. I installed a few. Then a few more. Then I copied skills from three different packs into one folder.

That is when things got weird.

Two skills had the same name. The agent picked one at random. Two more had almost identical descriptions, so both triggered on the same prompt. One skill had no description at all, so the agent never picked it. I had no idea until I started reading every SKILL.md by hand.

So I built a tiny tool to do that reading for me: skill-audit.

What it does

One command audits a whole skills folder:

skill-audit check --path ~/.claude/skills

It checks for:

Duplicate names — two skills declaring the same name
Invalid names — anything that is not kebab-case
Missing name / description — the description is how the agent decides when to trigger a skill
Missing frontmatter — a SKILL.md with no --- block
Duplicate / overlapping descriptions — identical text, or >=70% token overlap, so two skills fight over the same prompt
Description too short / too long — under 20 chars, or over 1024
Thin body — a header, not a skill
Missing SKILL.md — a folder agents will never discover
No version — so you can tell whether you have the latest

Exit code 0 means no errors. Warnings never fail the run on their own, so you can use it in CI.

Demo

$ skill-audit check --path examples/messy
[ERROR] duplicate-name: Duplicate skill name 'dup-skill'
[ERROR] invalid-name: 'Bad Name' must be kebab-case
[ERROR] missing-description: no-desc has no 'description'
[ERROR] missing-description: (unnamed) has no 'description'
[ERROR] missing-name: no-frontmatter/SKILL.md has no 'name'
[WARN ] frontmatter-parse: no-frontmatter/SKILL.md missing frontmatter block
[WARN ] no-version: (unnamed) has no 'version' in frontmatter
Summary: 5 error(s), 2 warning(s), 5 skill(s).

Try it

git clone https://github.com/techdevfun/skill-audit.git
cd skill-audit
PYTHONPATH=src python -m skill_audit check --path examples/good
PYTHONPATH=src python -m skill_audit check --path examples/messy || true

Zero dependencies, Python 3.9+ stdlib only. Tests + CI included.

Repo: https://github.com/techdevfun/skill-audit

Existing skill packs add more skills. This one tells you whether the ones you already have are healthy. If you maintain a pack, running this in CI catches the duplicate-name / missing-description breakage before your users do.

Your .env file is lying to you. I built a tiny tool to catch it.

TECHDEV — Fri, 16 Oct 2026 08:44:22 +0000

Your .env file is lying to you. I built a tiny tool to catch it.

Every new AI tool wants another API key. So my .env file kept growing — and one day I realized I no longer trusted it.

A teammate cloned the repo, followed .env.example, and the app crashed. A key that was in .env had never been added to the example. An old key that nobody used anymore was still there. And somewhere in the middle sat STRIPE_KEY=changeme, which had been there long enough that nobody noticed.

Formatters and linters check whether your .env file looks right. None of them told me whether my env setup was actually correct. So I built a small tool that does exactly that: env-doctor.

What it checks

One stdlib-only Python CLI. No dependencies, no network, no telemetry.

python -m env_doctor check --env .env --example .env.example

Check	What it finds
Drift	Keys in `.env.example` missing in `.env`, and vice versa
Duplicates	Same key defined twice
Placeholders	Empty values, `changeme`, `your-api-key`, `TODO`, `xxx`, ...
Risky secrets	Known prefixes (`ghp_`, `sk-`, `AKIA`, ...) or long high-entropy values
Unused vars	Keys not referenced in scanned code (`--scan .`)
Invalid lines	Lines without `=`, invalid key names

Exit code 0 means no errors, 1 means errors found. Warnings never fail the run on their own, so you can use it in CI without false failures from heuristics.

What it looks like

$ python -m env_doctor check --env examples/drift/.env --example examples/drift/.env.example
[ERROR] duplicate-key: Duplicate key 'DATABASE_URL' (line 3)
[ERROR] missing-in-env: 'NEW_API_URL' is in .env.example but missing in .env
[WARN ] placeholder-value: 'STRIPE_KEY' still looks like a placeholder value
[WARN ] missing-in-example: 'OLD_FLAG' is in .env but missing in .env.example
Summary: 2 error(s), 2 warning(s), 3 key(s).

That drift/ example ships in the repo, broken on purpose, so you can see the output immediately.

Try it in 3 steps

# 1. Clone
git clone https://github.com/techdevfun/env-doctor.git
cd env-doctor

# 2. No dependencies to install (Python 3.9+ stdlib only).
#    Optional: install the `env-doctor` command
pip install -e .

# 3. Run it
python -m env_doctor check --env .env --example .env.example

A few more ways to use it:

# Clean setup — should pass
python -m env_doctor check --env examples/basic/.env --example examples/basic/.env.example

# JSON output for CI / scripts
python -m env_doctor check --env examples/drift/.env --example examples/drift/.env.example --format json

# Also flag vars not referenced in code
python -m env_doctor check --env .env --example .env.example --scan .

And in CI, it's one step:

- name: Audit .env drift
  run: python -m env_doctor check --env .env --example .env.example --format text

The repo runs the same check in .github/workflows/ci.yml on Python 3.9, 3.11, and 3.12.

What's next

On the roadmap: --fix to auto-add missing keys to .env.example as empty placeholders, --strict to turn selected warnings into errors, support for .env.local / .env.production layering checks, and SARIF output for GitHub code scanning.

If you try it and it misses a case your project hits, open an issue — that's the feedback that shapes what gets built next.

Repo: https://github.com/techdevfun/env-doctor

DEV Community: TECHDEV

I installed a bunch of agent skills. Then I wrote a tool to audit them.

What it does

Demo

Try it

Your .env file is lying to you. I built a tiny tool to catch it.

Your .env file is lying to you. I built a tiny tool to catch it.

What it checks

What it looks like

Try it in 3 steps

What's next