- Home
- Skills
- Code Quality & Review
- Claims Ledger: Docs vs Code Audit with a CI Gate
Works with the AI Agents you already use
Claims Ledger: Docs vs Code Audit with a CI Gate
Audit project documentation against code and tests to generate a verified honesty registry and prevent doc drift.
$19
Claims Ledger: Docs vs Code Audit with a CI Gate
Example session with this skill installed
Audit the README.md and src/auth.ts to ensure our 'AES-256 encrypted' and 'Zero-knowledge' promises are actually implemented and tested. Use anchors, don't edit source files.
- Read your context and instructions
- Compiled the claims ledger
I have audited the authentication claims against the current implementation. I identified two atomic claims and traced them to the encryption logic and unit tests.
Claims Registry Audit
| ID | Statement | State | Label | Evidence | Gap |
|---|---|---|---|---|---|
| C-01 | AES-256 encrypted | enforced | PROVEN | auth.ts#L42, auth_test.ts#L12 | None |
| C-02 | Zero-knowledge | partial | PROVEN | auth.ts#L88 | No negative test for server-side key leakage |
Headline Findings
C-02 (Zero-knowledge) is marked partial. While the implementation uses client-side hashing, there is no test case verifying that the raw secret never reaches the POST /login payload.
- No literal overclaims (static badges) were found in the scanned files.
Wording Notes
- The README statement "Military-grade security" was dropped as it is pure opinion and non-falsifiable.
Next steps
- Review the
claims.jsonfile created at the root to verify the anchors are correct. - Add a test case in
auth_test.tsto exercise the zero-knowledge failure state to move C-02 toenforced. - Run
python3 scripts/claims_ledger.py renderto update the publicCLAIMS.mdfile.
Connects securely to your tools. The creator never sees your data.
About this skill
Document drift and marketing overreach create technical debt that erases user trust. This skill audits your project promises against the actual implementation, generating a machine-checked registry that links every claim to specific code and test files.
What it does
- Atomic claim extraction breaks down complex marketing copy or specs into individual, falsifiable statements.
- Evidence tracing maps every promise to its implementation mechanism and the specific tests that verify it.
- Epistemic labeling categorizes claims as proven, lemmas, or proposals to distinguish current reality from future roadmap.
- Overclaim detection scans UI and docs for literal verdicts like "PASS" or "secure" badges that aren't backed by computed logic.
- Gap reporting generates a structured registry of enforced, partial, and absent features with actionable fix instructions.
How it works
- Source analysis identifies promises in whitepapers, READMEs, landing pages, and UI text.
- Registry generation creates a
claims.jsonfile linking statements to source code anchors or trace comments. - Automated verification runs a checker script to validate that implementation files and passing tests exist for every
enforcedclaim. - Honesty rendering produces a
CLAIMS.mdtable that displays the gap for every partially implemented or absent feature.
Frameworks & tools
The skill uses a Python-based auditing script and a JSON registry schema. It integrates with standard testing suites by checking for test files and CI output.
Why this beats prompting it yourself
Generic prompts often miss the "fail-closed" logic required for real audits, leading to false positives. This skill enforces a strict evidence chain where a claim is marked absent by default unless specific code and passing tests are identified.
Use cases
- Due diligence preparing a technical implementation report for investors or auditors.
- Launch readiness ensuring that the landing page doesn't promise features that haven't merged yet.
- Compliance checking verifying that security badges shown in the UI are actually tied to backend checks.
- Drift prevention using a CI gate to fail builds when documentation changes without matching code updates.
Known limitations
Requires Python 3 to run the checker script. It cannot verify external third-party claims without local documentation and does not perform live runtime monitoring.
Requires Python 3. Verification is limited to static analysis of local code, tests, and documents. Cannot verify runtime performance without existing benchmark files.
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 14 days ago
Requires