Works with the AI Agents you already use
Verified Security Audit: Proven Findings, No False Alarms
Scanners and generic AI reviews often produce long lists of non-exploitable bugs.
$19
Verified Security Audit: Proven Findings, No False Alarms
Example session with this skill installed
Run a security review on the current PR. The app is a Node.js API deployed to the public internet. Check for injection and broken access control.
- Read your context and instructions
- Compiled the verified security audit
I have completed the security audit of the current PR changes. I mapped the attack surface across three new API endpoints and traced data flow to the database and filesystem.
VERIFIED SECURITY AUDIT: PR #42 (API Updates)
Scope: GET /api/v1/user/profile, POST /api/v1/reports/export | Deployment: public
VERIFIED (1)
[HIGH] Path Traversal in Report Export
- Source:
POST /api/v1/reports/export(any authenticated user) - Path:
routes/reports.ts:22->controllers/export.ts:45->utils/file.ts:12(fs.readFile) - Exploit:
{"format": "../../../.env"} - Proof:
test/security/export_traversal.test.tssuccessfully retrieved local environment variables. - Fix: Implement a strict allow-list for the
formatparameter and usepath.basename(). - Re-test: Exploit now returns a 400 Bad Request error.
TRACED, NOT REPRODUCED (0)
LEADS TO CHECK (1)
Potential IDOR in user profile update: The userId is taken from the request body at routes/user.ts:55, but I cannot confirm if the session middleware overwrites this before it reaches the service layer.
RULED OUT (2)
- SQL injection in
orders.tswas ruled out; query uses parameterizedWHEREclause. - Secrets in
config.tswere ruled out; values are loaded from environment variables at runtime.
WHAT WAS NOT REVIEWED
- Legacy
/adminroutes were outside the scope of this PR. - Third-party OAuth flow was not reviewed as it is handled by an external provider.
Next steps
- Review and apply the fix for the path traversal vulnerability in
utils/file.ts. - Run the provided regression test
test/security/export_traversal.test.tsto confirm the patch. - Verify if the
sessionmiddleware correctly handlesuserIdoverrides.
Connects securely to your tools. The creator never sees your data.
About this skill
Scanners and generic AI reviews often produce long lists of non-exploitable bugs. This noise causes teams to ignore security reports, allowing real vulnerabilities to slip through to production. This skill eliminates false positives by requiring every finding to be proven via a complete trace from an untrusted source to a dangerous sink.
What it does
- Vulnerability verification ensures every reported bug includes a documented path from input to sink.
- Exploit reproduction provides a safe, local test case or script to prove the vulnerability is real.
- Attack surface mapping identifies all entry points like HTTP routes, webhooks, and AI agent tools.
- Root cause fixing delivers minimal, targeted code changes and regression tests for every verified finding.
- Lead tracking separates suspicious patterns that lack a complete trace into a specific section for manual follow-up.
How it works
- Identify entry points and trust boundaries in the provided repository or pull request.
- Scan for dangerous sinks including database queries, shell calls, and unsafe file operations.
- Trace data flow backwards from the sink to the source, documenting every validation step along the path.
- Generate a proof using local reproduction scripts to verify the exploit works as expected.
- Produce the audit report classifying findings as Verified, Traced, or Leads with specific remediation diffs.
Frameworks & tools
This tool works with any codebase where data flows from external inputs to backend sinks, specifically targeting REST/GraphQL APIs, CLI tools, and AI agents with tool-calling capabilities.
Why this beats prompting it yourself
Generic prompts often hallucinate security flaws or flag harmless code, leading to wasted developer hours. This skill follows a rigorous trace-first methodology that forces the agent to prove reachability before reporting, ensuring every finding is actionable.
Use cases
- Reviewing a pull request for new injection vectors or access control flaws.
- Auditing an AI agent's tool definitions for prompt injection risks.
- Investigating if a known vulnerable dependency is actually reachable in your code.
- Performing a deep dive into authentication and session management logic.
Known limitations
Does not perform live penetration testing against production or shared staging environments. Findings that cannot be locally reproduced are marked as Traced or Leads.
Reproductions run only locally, never against production or shared environments. Without permission to run code, findings are marked traced, not reproduced. It reviews only the scope you give it and lists what it did not review.
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 18 days ago