Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Security Audit: Verified Vulnerabilities, Secrets & Fixes

    2

    Security scanners bury real risks under hundreds of false positives, and generic "review for security" prompts miss the issues that actually get exploited.

    Free

    14 uses5.0 (1 review)
    Security Audit: Verified Vulnerabilities, Secrets & Fixes

    Security Audit: Verified Vulnerabilities, Secrets & Fixes

    Example session with this skill installed

    Run a security audit on our API before the release. Check secrets, dependencies and auth.

    • Read your context and instructions
    • Compiled the security audit

    Example output on a sample Node.js (Express + Prisma) API.

    Security audit: api/ (full mode) — 4 findings

    Risk summary: 🔴 2 High · 🟡 1 Medium · 🔵 1 Low

    Attack surface: 23 routes (19 authenticated), 1 file upload, 2 outbound HTTP integrations, PostgreSQL via Prisma. Tools run: gitleaks, npm audit, semgrep p/owasp-top-ten.

    🔴 High — IDOR on order endpoint

    src/routes/orders.ts:42 · CWE-639
    Exploit: authenticated user B calls GET /orders/123 and receives user A's order (address, phone, items). The handler loads by id without an owner check.

    - const order = await db.order.findUnique({ where: { id } })
    + const order = await db.order.findFirst({ where: { id, userId: req.user.id } })
    + if (!order) return res.status(404).end()
    

    🔴 High — AWS access key in git history

    config/old.env (removed in commit a1b2c3, still in history) · CWE-798

    Action

    rotate the key first, then purge with git filter-repo --path config/old.env --invert-paths and add *.env to .gitignore.

    🟡 Medium — SSRF in webhook test endpoint

    src/routes/webhooks.ts:17 · CWE-918
    Exploit: POST /webhooks/test {"url":"http://169.254.169.254/latest/meta-data/"} reaches cloud metadata.
    Fix: allowlist schemes (https only), resolve DNS and block private/link-local ranges before fetching.

    🔵 Low — Verbose errors in production

    src/middleware/error.ts:9 returns stack traces. Return a generic message when NODE_ENV=production.

    Dependencies

    PackageVersionAdvisoryFixed in
    jsonwebtoken8.5.1GHSA-8cf7-32gw-wr339.0.0
    multer1.4.4GHSA-4pg4-qvpc-4q3h1.4.5-lts.1

    Hardening (no direct exploit)

    • Container runs as root: add USER node to the Dockerfile.
    • Add rate limiting to /auth/login.

    Next step: want me to apply fixes 1, 3 and 4 and bump the two packages, then re-run the tests?

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

    About this skill

    Security scanners bury real risks under hundreds of false positives, and generic "review for security" prompts miss the issues that actually get exploited. This skill audits like an application security engineer: **map the attack surface, trace untrusted input to dangerous sinks, prove exploitability, then fix.** ## What it does - **Maps the attack surface**: routes, handlers, auth middleware, file uploads, queue consumers, external calls and data stores. - **Runs the tools you already have**: gitleaks/trufflehog for secrets (including git history), npm/pnpm audit, pip-audit, govulncheck, cargo audit, and Semgrep OWASP rules when installed. Never installs anything silently; falls back to targeted git grep. - **Manual OWASP Top 10 review**: SQL/shell/HTML/NoSQL injection, IDOR and missing authorization, JWT and session flaws, SSRF, open redirects, path traversal, weak crypto, mass assignment, prototype pollution, data exposure in logs and errors. - **Checks config**: Dockerfiles, CI pipelines, CORS, debug flags, IAM/S3 policies. - **Verifies exploitability**: each candidate gets a concrete path (input → sink → impact) and is checked against existing guards. Non-exploitable items are dropped or moved to Hardening. - **Actionable report**: severity, CWE, file:line, exploit scenario and diff-style fix. Leaked secrets always come with "rotate first". ## Modes - **Diff mode**: audit only your branch before merging. - **Full mode**: whole repository, entry points first. ## Safe by design Read-only by default and never runs exploits against production or third-party systems. ## Best for - Pre-release and pre-pentest checks - Reviewing PRs that touch auth, payments or file handling - Startups without a dedicated security team

    • Static review plus local verification only; it is not a replacement for a full penetration test or runtime (DAST) testing.
    • Dependency results depend on the audit tools available in your environment and their advisory databases.
    • Business-logic flaws that require domain knowledge (e.g. pricing rules) are flagged as questions when intent is unclear.

    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

    5.0
    1 review
    5
    1
    4
    0
    3
    0
    2
    0
    1
    0

    This skill has been used 14 times.

    Trust & safety

    Security scan

    Security scan passed 11 days ago

    • Free with an Agensi account

    Requires

    Git

    Permissions

    Terminal / Shell
    Read Files

    Read Files: reads source, configs, Dockerfiles and CI workflows. Terminal: runs git and the security tools already installed in your environment (secret scan, dependency audit, SAST). Write Files: only if you ask it to apply fixes. It never runs exploits against production or third-party systems.

    Listed11 days ago
    Updated5 days ago

    What's inside

    Frequently Asked Questions

    No. It is a fast, developer-side audit that catches the most common exploitable issues (OWASP Top 10, secrets, vulnerable dependencies) before code ships. Use it before a pentest to fix the easy wins.

    No. Every candidate must have a concrete exploit path and is checked against existing guards. Non-exploitable items are dropped or listed separately as hardening tips.

    No. It uses gitleaks, trufflehog, semgrep and dependency auditors if they are installed, and falls back to targeted git searches and manual review otherwise. It never installs tools without asking.

    Only if you ask. It is read-only by default; when you ask for fixes it applies them one by one and re-runs your tests.

    Yes. Diff mode audits just your branch against the default branch, which is ideal before merging changes to auth, payments or file handling.

    New in Security