Skill Release Gate
Checks an Agent Skill bundle for structure, routing, safety and portability before release.
- Repository
- rohitg00/ai-engineering-from-scratch
- Stars
- 76,331
- Licence
- MIT
- Files
- 8 files
What it does
This skill makes an AI agent act as a pre-release reviewer for another agent skill (a folder with a SKILL.md instruction file). Before a skill is published, the agent reads the bundle's evaluation files, runs a bundled Python script that checks structure, trigger accuracy, artifact comparisons, safety findings, file-hash integrity and target-host compatibility, and reports the results layer by layer. The shipped path is a teaching fixture; the skill is explicit that a fixture pass is not a release claim.
Great for
- Maintainers of Agent Skill bundles who want a structured checklist before publishing
- Teams learning how skill evaluation, provenance and portability checks fit together
- Reviewers who need the exact script invocation and evidence recorded in a report
Skip it if
- Anyone expecting a one-click 'safe to publish' verdict: the shipped run is a deterministic fixture and the skill itself says a fixture pass never makes a release claim
- Production release decisions without captured evaluation results, an external trusted attestation and a separate trust channel, because the local checks only verify digests and cannot prove results were really captured
- Reviewing untrusted bundles on a machine with sensitive files, since the bundled script is executed locally against the bundle under review
Try asking
- “Run the release gate on my skill folder at ./my-skill in fixture demo mode and show me the exact argv first”
- “Check whether my skill's trigger cases have good precision and recall and list every raw observation”
- “Verify the file hashes in assets/manifest.json match my bundle and report any mismatch”
- “Summarise which portability targets in assets/hosts.json would drop required companion files”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 8 original files, the MIT licence and an attribution note.
skill-release-gate.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip skill-release-gate.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip skill-release-gate.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Start with the --fixture-demo path to see the report format, then do not treat its pass as a release decision
- Keep the trusted attestation and its hash outside the target bundle so the bundle cannot vouch for itself
- Read the failed-checks list rather than only the aggregate passed flag, since each layer is reported separately
What's inside
- assets/hosts.json
- assets/manifest.json
- evals/artifacts.json
- evals/cases.json
- evals/evidence.json
- references/eval-contract.md
- scripts/evaluate_skill.py
- SKILL.md
Redistributed unmodified from rohitg00/ai-engineering-from-scratch under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.