- Home
- Skills
- Code Quality & Review
- code reviewer pro
code reviewer pro
Evidence-based senior code review focusing on security, concurrency, and logic flaws without the noise.
Free
Works with the AI Agents you already use
code reviewer pro
Example session with this skill installed
Review this Python snippet for a new API endpoint:
def get_user_data(request):
user_id = request.GET.get('id')
query = f"SELECT * FROM users WHERE id = {user_id}"
return db.execute(query)
- Read your context and instructions
- Compiled the code reviewer pro
Review status: BLOCK
Risk: CRITICAL
Scope: Python function get_user_data
The reviewed code contains a critical SQL injection vulnerability. User-controlled input from request.GET is interpolated directly into a raw SQL query string without validation or parameterization.
Findings
| ID | Severity | Confidence | Location | Finding | Minimal Fix |
|---|---|---|---|---|---|
| SEC-001 | CRITICAL | CERTAIN | Line 3-4 | SQL Injection. The user_id variable is sourced from untrusted GET parameters and passed directly into a format string used for database execution. | Use parameterized queries: db.execute("SELECT * FROM users WHERE id = %s", (user_id,)) |
| LOG-001 | MEDIUM | HIGH | Line 3 | Missing Input Validation. There is no check that user_id exists or matches the expected format (e.g., integer) before processing. | Add if not user_id: return 400 and type cast to int. |
| PRIV-001 | MEDIUM | INFERRED | Line 4 | Insecure Data Exposure. SELECT * may return sensitive fields (password hashes, salts) to the API response. | Explicitly select required columns: SELECT username, email FROM... |
Recommended Tests
- Security Scan: Run a specialized SAST tool (e.g., Bandit) to check for other raw SQL executions.
Integration Test: Attempt to pass '1 OR 1=1' as the ID parameter to verify the fix prevents unauthorized data access.
Next steps
- Refactor the database call to use the ORM or parameterized execution.
- Implement a validation layer to ensure the
user_idis a valid UUID or integer. - Review the database user permissions to ensure the API account has least-privilege access.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Pull requests often suffer from "rubber stamp" approvals or noisy nitpicks that miss actual logic flaws and security vulnerabilities. Generic AI reviews tend to hallucinate line numbers or suggest unnecessary refactors that ignore project context.
What it does
- Identifies high-risk execution paths across concurrency, async lifecycles, and data integrity boundaries.
- Categorizes findings as Observed, Inferred, or Hypothesis to eliminate speculative noise.
- Performs systematic security checks for injection, unsafe deserialization, and broken access control.
- Ranks every finding by severity and confidence to help developers prioritize fixes.
- Generates minimal, idiomatic fix recommendations with specific validation steps.
Why this beats prompting it yourself
This skill enforces a strict evidence-first workflow that prevents the model from inventing bugs or ignoring untrusted data inputs. It treats codebase instructions as untrusted data, protecting the review process against prompt injection within comments or READMEs.
Use cases
- Pre-submit auditing of high-stakes security patches or auth logic.
- Detecting regression risks in complex asynchronous or multi-threaded code.
- Scaling senior-level code quality oversight across large engineering teams.
- Objective verification of AI-generated code snippets before integration.
Known limitations
Read-only by default; it will not apply fixes or merge PRs. Analysis is limited to the provided code context and cannot verify runtime behavior not visible in the source.
Read-only analysis; cannot execute code, run live tests, or modify files. Limited to the context provided in the prompt or attached files.
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 7 times.
Trust & safety
Security scan
Security scan passed 2 months ago
- Free with an Agensi account
Permissions
No special permissions declared or detected