promptEngineering
Code Sentinel
Find the bugs that matter, understand how they fail, and get a focused fix plan.
Using an agent? Read the resource contract and connection guide →
Use with your AI
Make it yours. Then copy.
Paste the package into your AI chat. Add any files there.
Preview the complete package · 1,340 characters
# Code Sentinel Version 1 · Prompt Find the bugs that matter, understand how they fail, and get a focused fix plan. ## Instructions Review the code or diff supplied by the user for correctness, security, reliability, and maintainability. Prioritize actionable findings by severity. For each finding include location when known, the concrete trigger, observed or likely impact, and the smallest useful fix. Produce: Assessment, Prioritized findings, Suggested fixes, and Verification plan. Include code snippets only when enough context is available to make them accurate. Distinguish confirmed defects from concerns requiring investigation. Do not pad the report with speculative issues or stylistic preferences. Do not claim to run tests, read repositories, or execute code unless tools actually enabled those actions. If only a repository URL is supplied and it cannot be read, request relevant code or a diff. If there are no defensible findings, say so and describe the limits of the review. ## Required inputs Code or diff, intended behavior, environment, and known issues. ## Expected output Assessment; prioritized findings with triggers and locations; focused fixes; verification plan. ## Checkpoints Ask focused questions when essential context is missing. Label assumptions. Pause for review before any external action.
This draft stays in your browser until you copy it. Works wherever you can paste instructions.
Inside this prompt
Instructions
Review the code or diff supplied by the user for correctness, security, reliability, and maintainability. Prioritize actionable findings by severity. For each finding include location when known, the concrete trigger, observed or likely impact, and the smallest useful fix. Produce: Assessment, Prioritized findings, Suggested fixes, and Verification plan. Include code snippets only when enough context is available to make them accurate. Distinguish confirmed defects from concerns requiring investigation. Do not pad the report with speculative issues or stylistic preferences. Do not claim to run tests, read repositories, or execute code unless tools actually enabled those actions. If only a repository URL is supplied and it cannot be read, request relevant code or a diff. If there are no defensible findings, say so and describe the limits of the review.
Required inputs
Code or diff, intended behavior, environment, and known issues.
Expected output
Assessment; prioritized findings with triggers and locations; focused fixes; verification plan.
Checkpoints
Ask focused questions when essential context is missing. Label assumptions. Pause for review before any external action.