Fix a bug with root-cause analysis
Makes the agent find the cause before patching symptoms.
Agent task · Cursor / coding agents, Claude Code, ChatGPT +1
Library / Coding & Technology / Code review
Prioritised review: bugs, security, tests, then style.
Review this diff: """ [paste diff or say: review the current git diff] """ Report findings in this order, skipping empty categories: 1. Bugs / incorrect logic 2. Security (input validation, authz, secrets, injection) 3. Performance problems (N+1 queries, unbounded loops, missing indexes) 4. Missing tests 5. Readability (only if significant) For each: file:line, the problem, why it matters, a concrete fix. End with an overall verdict: approve / approve with changes / request changes.
Copies your prompt and opens the official site in a new tab. You may need to paste it yourself (Ctrl/⌘ + V). Nothing is sent to the AI by us.
Replace [bracketed] parts with your own details before sending. Want it tailored? Build a custom coding assistant prompt.
A typical one-line request compared with this prompt. We only show sample responses where we've written an illustration.
Prompt
code review
2 words · no role, context, format or constraints
Prompt
Review this diff: """ [paste diff or say: review the current git diff] """ Report findings in this order, skipping empty categories: 1. Bugs / incorrect logic 2. Security (input validation, authz, secrets, injection) 3. Performance problems (N+1 queries, unbounded loops, missing indexes) 4. Missing tests 5. Readability (only if significant) For each: file:line, the problem, why it matters, a concrete fix. End with an overall verdict: approve / approve with changes / request changes.
75 words · generated instantly from templates, no AI called
What we added
Makes the agent find the cause before patching symptoms.
Agent task · Cursor / coding agents, Claude Code, ChatGPT +1
Behaviour-preserving refactor with tests as a safety net.
Agent task · Cursor / coding agents, Claude Code, ChatGPT +1
Covers happy path, edge cases and failures without testing implementation details.
Agent task · Cursor / coding agents, Claude Code, ChatGPT +1