- Home
- Skills
- Prompt & Skill Engineering
- Self-Audit
Works with the AI Agents you already use
Self-Audit
Inward adversarial self-audit for any agent workflow — catches the mistakes your checklist cannot.
$9
/moSubscription · cancel anytime
Secure checkout via Stripe
- Always the latest version
- Delivered live to your agent via MCP
- Cancel anytime, access ends at period end
Self-Audit
Example session with this skill installed
Run a self-audit on the documentation generation workflow. Check for any contradictions in our 'minimalist' principle versus the current output volume and look for missing gates.
- Read your context and instructions
- Compiled the self-audit
Audit complete for the documentation generation workflow. I identified one contradiction regarding output volume and two missing gates in the markdown cleanup phase.
Audit Report: 2024-05-20
Inward Adversarial Review
Vector C (Contradiction): FOUND. The "No Bloat" principle conflicts with the "Comprehensive API Coverage" rule. Applying both to internal helper methods creates ambiguity.
- Vector D (Absence): FOUND. Step 4 (Summary Generation) makes a stylistic choice but has no quality gate or veto.
Scorecard
| Dimension | Status | Evidence |
|---|---|---|
| D1: Script Validation | PASS | verify_md.sh exited with code 0. |
| D2: Derived Fact Drift | FAIL | entity_count in README (12) differs from source (14). |
| D3: Gate Adherence | WARN | Summary generated without secondary verification. |
| D4: Clean Workspace | PASS | No untracked files found in build directory. |
Lessons Learned
- L004: Recompute
entity_countdirectly from/src/definitionsto prevent README drift. (Status: Active) - L005: Implement a D0 gate for helper method documentation to resolve bloat contradictions. (Status: Active)
Next steps
- Approve the proposal in
proposals/2024-05-20.mdto add a validation script for entity counts. - Review the re-scoped "No Bloat" definition in the project guidelines.
- Integrate the new summary-veto step into the main workflow.
Subscription · always the latest version
- Secure checkout via Stripe
- Cancel anytime
- Delivered live via MCP
About this skill
Most automations only check the actions already written into their checklist. This skill audits the checklist itself. It runs an inward adversarial review with two scans — a contradiction scan (do two of your own rules fight?) and an absence scan (is there a gate where there should be one?) — plus a three-question method to derive your own scorecard's dimensions, and a lesson loop that turns repeatable mistakes into new checks. It reviews the process, not the content, and never waits for a human to point out a problem.
Focuses on process integrity over content accuracy. Requires filesystem access for persistent lesson memory. Vectors C and D require LLM judgment and cannot be fully automated via simple scripts.
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
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 5 days ago