Cursor
Usage guidelines
Practical ways to get value from Cursor while staying safe and reviewable
Use it for the right tasks
- Codebase understanding: ask for a high-level map (modules, data flow, key entry points).
- Small, safe changes: ask for focused diffs rather than broad rewrites.
- Refactors: ask for a staged plan and do it in small PRs.
- Tests: ask it to propose test cases first, then implement.
- Docs: ask for concise docs that match existing conventions.
Ask for outputs you can verify
Good prompts typically include:
- The goal and non-goals
- Constraints (performance, compatibility, style conventions)
- Where to change (specific folders/files)
- What “done” means (tests passing, no behavior change, migration plan, etc.)
Keep changes reviewable
- Prefer incremental commits/PRs.
- Avoid “mega diffs” that mix formatting, refactors, and feature changes.
- Ask Cursor to summarize risks and edge cases before you submit.
Security and confidentiality
Cursor is a third-party platform. Follow the company policy and secure usage principles:
- Disable / opt out of any available data sharing / model improvement / training settings wherever technically possible.
- Never share production credentials, API keys, private certificates, passwords, or database connection strings.
- Avoid sharing confidential client information unless absolutely necessary and contractually permitted.
- Prefer sanitized, redacted, or masked data and share the minimum code needed for the task.
- If in doubt, escalate to your Team Leader / Reporting Manager before usage.
Recommended workflow
- Start by asking for a plan (steps + impacted files).
- Implement in small chunks and run tests.
- Ask for a review checklist (what to verify before PR).
- Treat AI output as a draft; refine it to meet Infynno quality expectations.