the quality gate for Agent Skills
A skill's description is the only signal an agent uses to decide whether to invoke it —
routing logic, not documentation. Tripwire lints your SKILL.md, probes activation with
real agent sessions, and gates every PR.
npm install -g tripwire-skills
from a scan of real published SKILL.md files ·
methodology ↗
no install needed
The real lint engine, bundled to your browser — nothing leaves this page. Full activation probing runs from the CLI and the Action.
three commands
Static checks are instant and free. Activation probing runs real agent sessions. Committed scenarios make both a regression gate.
Instant, offline checks: description starts with "Use when", stays under 1024 chars, avoids workflow summaries; name is kebab-case; the body has real examples and isn't a stub.
Generates a prompt matrix across four zones and runs real agent sessions to map what fires, what silently misses, and what false-triggers — the check a linter can't do.
Commit the generated scenarios and rerun them deterministically — in CI or locally — to catch activation regressions before they ship.
the part a linter can't catch
Every generated prompt lands in one of four zones. The report tells you exactly where your skill over- or under-fires — and what to change.
runs on every pull request
Gate skill changes the way you gate lint and tests. One workflow file — your API key stays on your runner.
SKILL.mdFork PR without a key? The probe skips with a notice — lint still gates the PR.
# gate skill changes like you gate tests name: Tripwire on: pull_request jobs: skills: runs-on: ubuntu-latest permissions: { contents: read, pull-requests: write } steps: - uses: actions/checkout@v5 with: { fetch-depth: 0 } - uses: bharath31/tripwire@v1 with: probe: true anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}
as your skills grow
Two skills that each lint clean can still fight at runtime. Scan a whole directory for name collisions and description overlap — skills shadowing each other.
Did the skill actually do the right thing, not just fire? Author assertions plus an optional LLM-judge rubric per case.
A scenarios file that passed in March can silently regress in June. Re-probe on a schedule and fail loudly when activation behavior drifts.
Claude Code is verified; experimental Gemini CLI and Codex adapters let you compare activation behavior from the same SKILL.md.
extends: tripwire:recommended, ESLint-style. Turn rules off, change severity, or add org-specific checks as plain JS — applied everywhere, including the Action.
A live README coverage badge, plus one-command fixes for mechanically-safe lint issues. Tripwire also ships as a skill and an early VS Code extension.
production questions
A linter checks file structure and authoring rules. Tripwire also runs the skill through a real agent and observes whether it activates. That behavioral check catches valid-looking descriptions that miss intended prompts or trigger for unrelated requests.
Tripwire does not send skill files, prompts, file paths, repository names, model output, usernames, or credentials to its servers. Probes run locally or on your GitHub runner through your configured agent provider. Anonymous behavioral telemetry can be disabled with TRIPWIRE_TELEMETRY=0.
npm install -g tripwire-skills