Works with the AI Agents you already use
Pre-Release Audit: GO / NO-GO check for code and AI skills
Pre-Release Audit checks your project right before you ship and gives one clear verdict: GO, GO WITH WARNINGS, or NO-GO.
$9.99
Pre-Release Audit: GO / NO-GO check for code and AI skills
Example session with this skill installed
Audit this skill before I publish it. Use the pre-release-audit skill.
- Read your context and instructions
- Compiled the pre-release audit
Summary
Audited a skill package (SKILL.md + 9 files, 16.9 KB) before publishing.
Verdict: NO-GO. 1 blocker, 5 warnings.
RELEASE VERDICT: NO-GO
Mode: AI Skill Audit
Detected: skill package (SKILL.md, 9 files, 16.9 KB)
Blocker
SKILL.mdline 3: top-levelversionkey is not allowed; validators reject the skill.
Fix: delete line 3 and addmetadata:withversion: "2.0.0".
Warnings
SKILL.mdline 4: description has no "use this when" triggers, so the agent may not pick the skill.LICENSE.txt: no copyright line or year.- Zip is wrapped in
skills/<name>/instead of<name>/at the root. - The skill's rule and its own test disagree on the number of tags.
- No test log; tests were not run in the real client.
Checklist
| Area | Result | Note |
|---|---|---|
| S1 Package structure | WARN | extra wrapper folder |
| S2 Frontmatter | FAIL | disallowed key: version |
| S3 Description quality | WARN | no trigger wording |
| S5 Safety rules | PASS | |
| S6 Secrets and paths | PASS | 0 hits in 9 files |
| S7 License and IP | WARN | no copyright line |
| S9 Test evidence | WARN | MANUAL: not run in client |
Next steps
- Fix the blocker, then run the audit again.
- Ask the agent to apply the fixes; it waits for your yes before changing anything.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop finding release bugs after your users do.
Pre-Release Audit checks your project right before you ship and gives one clear verdict: GO, GO WITH WARNINGS, or NO-GO. Every problem comes with a file location and a concrete fix.
Just tell your agent: “Is this ready to release?”, “Audit before I tag v2.1”, or “Check my skill before I publish it.”
Two modes, one skill
Code Release mode first detects what your project is, then runs only the checks that apply:
Detects: Node and web frameworks, Python, Go, Rust, Java/Kotlin, .NET, PHP, Ruby, Android/APK, Flutter, React Native, iOS, Docker, desktop EXE apps (Electron, Tauri, PyInstaller), static sites, and monorepos (each package audited separately). Checks: git state, version and changelog, committed secrets and keystores, debug leftovers, lockfiles and dependencies, tests and build, destructive migrations, missing environment variables, Docker and CI config, Android release settings, desktop installer settings. If it can’t identify the type, it says “Type not detected” and runs only the generic checks instead of guessing.
AI Skill Audit mode checks a SKILL.md package before you publish it:
Package structure: the zip layout, missing files, and junk files. Frontmatter: invalid keys and name rules. Description quality: whether it will actually trigger. Body quality: contradictions and instruction clarity. Safety and honesty rules: no invented data, no guarantees, no credential requests. Secrets and personal paths: leaked keys, emails, and local paths. License and copyright line. Version consistency and claims versus what the skill really does. Test evidence: whether tests were run in a real client. What you get
A short report: verdict first, then blockers (with fixes), warnings, a pass/fail table, what could not be checked, and a rollback note.
Built to be safe Read-only. It never edits files, commits, tags, pushes, or deploys. Never prints your secrets. It shows only the location and a redacted preview. Honest. If it can’t verify something, it says SKIPPED or MANUAL, never a false PASS. No installs. It uses the test and build commands your project already has. Good for
Solo developers, small teams, indie app makers, and anyone publishing skills or tools to a marketplace without a dedicated release engineer.
Not for
It doesn’t replace a full security audit or penetration test, and it can’t run your tests in your real client for you. It’s a fast pre-ship safety check.
Read-only. Requires git for version diffs. Won't install new tools. Maximum 60-line report limit. Skips deep security audits in favor of fast release checks.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 7 days ago
Requires
Permissions
Read-only audit: it never edits files, commits, pushes or deploys. It reads project files to find secrets, version and config problems, and runs git plus your project's own test, build and audit commands. Dependency audits contact package registries. It asks first before slow builds.