Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    AI Code Security Reviewer — Audit Agent-Written Code

    1

    Your code reviewer assumes the author is human. AI code fails differently: hallucinated packages, invented APIs, removed guardrails. Catch all three.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    AI Code Security Reviewer — Audit Agent-Written Code

    AI Code Security Reviewer — Audit Agent-Written Code

    Example session with this skill installed

    Review this Flask endpoint my agent generated — user lookup plus an email send. Here's app.py, config.py, requirements.txt.

    • Read your context and instructions
    • Compiled the ai code security

    AI Code Security Review — lookup_endpoint.py — verdict: DO NOT MERGE

    | ID | Sev | Category | Issue |
    | ACR-1 | CRITICAL | hallucinated-dep | flask-secure-headers doesn't exist on PyPI — slopsquat risk |
    | ACR-2 | CRITICAL | injection | f-string SQL in email lookup |
    | ACR-3 | HIGH | fabricated-api | jwt.decode(verify=False) — param invented in PyJWT 2.x |
    | ACR-4 | HIGH | secrets | SECRET_KEY = "dev-secret-123" shipped |
    | ACR-5 | MEDIUM | guardrail-removal | rate_limit decorator dropped in refactor |

    Deps: 4 checked | 1 fake | 1 invented API. Coverage: 5/5 passes, 112 lines. VERIFY-EXTERNAL items listed.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Your code reviewer assumes a human wrote the code. AI-written code fails differently — confidently, silently, in patterns humans rarely produce. This skill hunts those patterns: hallucinated dependencies (slopsquatting), fabricated API calls, silently removed guardrails, placeholder secrets shipped as real, and OWASP-class bugs.

    What it does

    • Dependency verification table — every import checked: does the package exist? Is the API real? Is it pinned? Unverifiable items flagged VERIFY-EXTERNAL, never guessed.
    • AI-slop bug hunt — cargo-cult crypto (md5(password), jwt.decode(verify=False)), theater validation (checks that run but aren't enforced), debug flags and example keys shipped to prod.
    • Guardrail diff — enumerates safety code the agent deleted or weakened: rate limits, auth checks, allowlists, validation.
    • OWASP pass — injection, authn/authz gaps, secrets, SSRF, traversal, XSS, wildcard perms.
    • Merge verdict — SAFE TO MERGE / MERGE WITH FIXES / DO NOT MERGE, with quoted evidence for every finding. A finding that can't point at real code doesn't ship.

    Why it beats generic review

    "Review this code" gives you style notes and a confident nod. This gives you the answer to the actual question: did the model get it right? Fake packages get caught before pip install does the attacker's work, and removed guardrails get caught before the refactor merges.

    Known limitations

    Static review of provided code only — no execution, no registry lookups. Complements SAST in CI; doesn't replace it.

    Static review of provided code only — no execution, no registry lookups (unverifiable items are flagged VERIFY-EXTERNAL, not guessed). Complements, not replaces, SAST tooling in CI.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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 10 days ago

    Permissions

    No special permissions declared or detected

    Listed10 days ago

    What's inside

    Frequently Asked Questions

    Generic review checks correctness and style. This targets LLM-specific failure modes: packages that don't exist (slopsquatting), API calls the library never had, guardrails removed during 'refactors', and placeholder secrets shipped as real.

    No — it works from provided code and marks anything it can't confirm as VERIFY-EXTERNAL with the exact check to run, rather than pretending it verified a registry.

    New in Security