- Home
- Skills
- Prompt & Skill Engineering
- Skill Gate Auditor — Pre-Flight Review for Skill Packages
Works with the AI Agents you already use
Skill Gate Auditor — Pre-Flight Review for Skill Packages
Professional audit tool to verify AI skills against marketplace security, honesty, and quality standards.
$13
Skill Gate Auditor — Pre-Flight Review for Skill Packages
Example session with this skill installed
Audit the attached skill-repo-cleaner package. It includes SKILL.md and a cleanup.sh script. Check for security and instruction clarity.
- Read your context and instructions
- Compiled the gate auditor
I have completed the procedural audit of the skill-repo-cleaner package. I performed 3 grep searches for obfuscation, network calls, and secrets, finding one undeclared network hit in the shell script.
Skill audit: skill-repo-cleaner
| # | Criterion | Verdict | Evidence |
|---|---|---|---|
| 1 | Frontmatter & filenames | PASS | SKILL.md:1-5 |
| 2 | Honest description | PASS | Description matches cleanup.sh logic |
| 3 | Hidden behavior | PASS | No base64 or injection patterns found |
| 4 | Network calls | FAIL | cleanup.sh:12 contains curl -X POST |
| 5 | Credentials & secrets | PASS | No API keys or tokens detected |
| 6 | Originality | PASS | Logic appears unique to this implementation |
| 7 | Guess-free instructions | REVISE | SKILL.md:18 refers to "config" but file is missing |
| 8 | Scripts do what they say | FAIL | cleanup.sh sends telemetry via curl |
Patches
- [4] cleanup.sh:12 — Delete the line
curl -X POST https://telemetry.dev/log...as it is undeclared and unnecessary for local cleanup. - [7] SKILL.md:18 — Replace "Load settings from config" with "Use default settings defined in the cleanup.sh header" to remove reliance on a missing file.
- [8] cleanup.sh:1-24 — Remove the telemetry block to align script behavior with the description "performs local-only file cleanup".
Overall: FAIL
Next steps
- Remove the telemetry
curlcall fromcleanup.shto fix the security violation. - Update the instructions in
SKILL.mdto remove the reference to the non-existent config file. - Rerun the audit once these patches are applied to move to a PASS verdict.
Connects securely to your tools. The creator never sees your data.
About this skill
Developing a skill for an AI marketplace requires following strict review standards to avoid rejection. This skill audits your package against the precise criteria marketplaces use to evaluate quality, security, and integrity. It scans for hidden network calls, obfuscated code, and credential harvesting while ensuring your instructions are actually executable.
What it does
- Structure verification checks frontmatter, kebab-case naming, and file inventory consistency.
- Security scanning identifies base64 blobs, undeclared network calls, and hardcoded API secrets.
- Instruction auditing ensures every step is guess-free with concrete inputs and no vague references.
- Capability mapping verifies that all claims made in the description are fully supported by the instructions.
- Script analysis line-by-line identifies side effects to ensure scripts match their stated purpose.
How it works
- Point the auditor at a skill directory containing a
SKILL.mdand any associated scripts or data files. - The agent performs a procedural 8-step audit covering frontmatter, honesty, obfuscation, network, secrets, originality, logic, and side effects.
- Review the generated audit table which marks each criterion as PASS, REVISE, or FAIL with specific line references.
- Apply the provided concrete patches to fix REVISE or FAIL verdicts before submission.
Frameworks & tools
Works with any skill package consisting of Markdown files, shell scripts, Python, or JavaScript. It uses read-only file analysis and pattern matching.
Why this beats prompting it yourself
Generic prompts often overlook subtle security risks like undeclared fetch calls or base64 prompt-injection patterns. This skill enforces a procedural 8-point checklist that demands file:line evidence for every claim, preventing hallucinations and vague feedback.
Use cases
- Audit your own skill package before submitting to a public marketplace to ensure a first-time pass.
- Review a third-party skill downloaded from the web before giving it access to your local environment.
- Debug why a submitted skill was rejected when the marketplace feedback is generic or unclear.
- Standardize quality control for a team of developers building internal agent tools.
Known limitations
This is a procedural audit based on file content and does not execute scripts in a sandbox. It cannot verify if a skill is "fun" or "useful," only if it is safe, honest, and technically sound.
Does not sandbox scripts. Cannot verify external API uptime. Limited to static analysis of files within the provided package.
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 20 days ago