How does this security review agent differ from a standard linter or static analysis tool?

This skill specifically focuses on evidence-based data flow analysis and the unique risks of AI tool-calling, such as prompt injection and excessive tool permissions, which standard linters often miss.

Which AI agents and development environments is this skill compatible with?

The skill is designed to work with any AI agent capable of file-system access or Git integration, and it generates output in standardized formats compatible with most developer workflows.

How does the agent handle false positives and ensure the findings are actionable?

The agent provides a confidence score for every finding, distinguishing between confirmed vulnerabilities and potential risks that require manual verification, significantly reducing false positives.

What is required to set up and integrate this skill into my current CI/CD pipeline?

Setup involves granting your host agent access to your codebase or PR diffs and configuring the security review skill as a specialized task handler within your existing automation pipeline.

Does the security review agent require execution access to my live production infrastructure?

No, this skill performs a read-only analysis of the source code and configuration files you provide; it does not execute code or interact with your live production environment.

ai security review agent

Example session with this skill installed

Review this Node.js API handler for security risks:

app.get('/api/user/data', async (req, res) => {
  const user = await db.users.findOne({ id: req.query.userId });
  res.json(user);
});

Read your context and instructions
Compiled the ai security review

Security Review Summary

Scope: Node.js API handler (/api/user/data)
Review status: Completed
Release decision: BLOCK

1. Executive Findings

HIGH: Broken Object Level Authorization (BOLA) - Confirmed. The handler fetches user data based on a client-supplied ID without verifying ownership.

2. Attack Surface and Trust Boundaries

Input: req.query.userId (Untrusted)
Trust Boundary: API Endpoint -> Database Query
Privileged Operation: Reading sensitive user records from db.users.

3. Findings

[SEC-001] Missing Authorization Check on User Data Access
Severity: HIGH
Confidence: Confirmed
Location: app.get('/api/user/data', ...)
Evidence: The code uses req.query.userId directly in a database lookup and returns the result without checking if the authenticated requester has permission to view that specific ID.
Attack Path: Authenticated Attacker -> Provide victim's userId in query string -> Server fetches victim's data -> Data returned to attacker.
Impact: Unauthorized disclosure of sensitive user information across the entire user base.
Remediation: Validate that the userId in the query matches the id of the currently authenticated session user.
Verification:

Attempt to access /api/user/data?userId=[VICTIM_ID] using an attacker's session; confirm 403 Forbidden.
Access /api/user/data?userId=[MY_ID] using own session; confirm 200 OK.

4. Security Gaps / UNKNOWNs

Authentication Middleware: The provided snippet does not show how req.user is populated or if an auth middleware is present.

5. Verification Plan

Test Case 1: IDOR/BOLA attempt using mismatched session and query IDs.
Test Case 2: Success path for legitimate owner access.

6. Release Gate

BLOCK
The BOLA vulnerability is confirmed and allows unauthorized data access. The code must not be deployed until ownership verification is implemented.

Next steps

Implement ownership validation using the authenticated session context.
Provide the authentication middleware code for a follow-up review.
Run the suggested verification tests in the staging environment.

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

About this skill

The problem

Code reviews often miss subtle authorization gaps, injection vectors, or insecure AI tool-calling logic. Developers frequently rely on generic linters that lack the context to trace untrusted data across trust boundaries.

What it does

Analyzes pull requests, infrastructure diffs, and dependency manifests for evidence-based security weaknesses.
Traces controllable inputs from entry points like API handlers to privileged operations or database queries.
Audits AI-specific risks including prompt injection, instruction confusion, and excessive tool permissions.
Classifies findings by severity (Critical to Info) and confidence level, distinguishing confirmed bugs from hypotheses.
Generates targeted remediation advice and specific verification tests to block abuse paths.

Why this beats prompting it yourself

Generic prompts often hallucinate vulnerabilities or miss complex attack paths like SSRF and insecure deserialization. This skill enforces a strict evidence-based review procedure that separates confirmed facts from assumptions, ensuring you don't waste time on false positives.

Use cases

Reviewing PRs for Broken Object Level Authorization (BOLA) and injection risks.
Auditing LLM-integrated applications for prompt injection and unsafe tool execution.
Checking infrastructure-as-code and dependency changes for misconfigurations or leaked secrets.
Providing a formal security release gate for compliance or internal audit trails.

Known limitations

Read-only analysis based strictly on provided material. It cannot verify runtime behavior, external services, or infrastructure not explicitly defined in the input files.

How to use

Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

Reviews

No reviews yet

Be one of the first to try it. Every listed skill passes our trust checks below.

Security scanned

Passed our 8-point scan before listing

1 install

Downloaded by developers to date

30-day refund

Not a fit? Get your money back

ai security review agent