Security Code Review
Automated security code reviewer that finds logic-level vulnerabilities and provides severity-rated fixes.
Free
Works with the AI Agents you already use
Security Code Review
Example session with this skill installed
Review the security of this diff: `app.get('/api/report/:id', (req, res) => { const query = "SELECT * FROM reports WHERE id = " + req.params.id; db.execute(query, (err, row) => { res.send(row); }); });`
- Read your context and instructions
- Compiled the security code review
The security review is complete. I found two high-severity issues in the provided diff. ## Security Review Results | Severity | Category | CWE | Location | Finding | Fix | | :--- | :--- | :--- | :--- | :--- | :--- | | **Critical** | Injection | CWE-89 | Line 2 | Direct string concatenation of `req.params.id` into a SQL query allows for SQL injection. | Use parameterized queries: `db.execute("SELECT * FROM reports WHERE id = ?", [req.params.id], ...)` | | **High** | Broken Access Control | CWE-639 | Line 2 | IDOR: The endpoint fetches a report by ID without verifying if the authenticated user has permission to view that specific report. | Add an ownership check: `WHERE id = ? AND owner_id = ?` using the session user ID. | ### Verified Categories * **SSRF**: Checked; no external fetches performed in this hunk. * **Path Traversal**: Checked; no file system operations identified. * **Insecure Deserialization**: Checked; input is treated as a string/integer, not an object. **Verdict: Request Changes** The SQL injection vulnerability is directly exploitable. Please implement parameterization and add authorization checks before merging. **Next steps** 1. Refactor the database call to use the driver's parameterization syntax. 2. Implement a middleware check to validate user ownership of the requested `:id`. 3. Run a full-repo secret scan if these credentials were ever committed to a public branch.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Static analysis tools often miss logic-level vulnerabilities like broken authorization or business-logic bypasses. Developers often merge PRs with hidden security flaws because they lack the time for a deep, adversarial review of every change.
What it does
- Analyzes diffs and pull requests for OWASP Top 10 and CWE Top 25 vulnerabilities.
- Identifies logic-level flaws such as IDOR, SSRF, and unsafe trust boundary crossings.
- Maps every finding to a specific CWE and provides a concrete, language-specific code fix.
- Assigns severity ratings based on exploitability to reduce false-positive fatigue.
- Delivers a final verdict (Approve/Request Changes) suitable for CI/CD integration.
Frameworks & tools
Supports JS/TS, Python, Go, Java, PHP, Ruby, and Rust. Compatible with git environments for diff and context analysis.
Why this beats prompting it yourself
This skill establishes trust boundaries before scanning, preventing the "hallucination fatigue" of generic AI reviews. It follows a systematic checklist across vulnerability classes and mandates a specific severity rubric, ensuring your security standards remain consistent across the entire codebase.
Use cases
- Reviewing a pull request for vulnerabilities before merging into main.
- Automating security gates in CI/CD pipelines to block high-severity flaws.
- Auditing a specific file or endpoint for exploitable logic bugs.
- Validating a fix for a previously identified security issue.
Known limitations
Does not replace dedicated dependency/CVE scanners or secret-scanning tools. It performs static review and does not execute code for dynamic testing.
Not a replacement for secret scanners or dependency CVE tools. Static review only; no dynamic execution. Performs best when repo access is available for surrounding context.
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
This skill has been used 34 times.
Trust & safety
Security scan
Security scan passed 2 months ago
- Free with an Agensi account