AI prompt library | DevPrompt Lab

Published prompts

• Analyze cache invalidation and freshness

  Map cache keys, stale-data risks, and invalidation triggers before optimizing reads.

• Choose instruction files for coding tools

  Map project context into each tool's documented instruction format without inventing support.

• Create a risk-based test plan

  Choose focused tests that protect behavior and important failure cases.

• Debug with a reproducible hypothesis

  Turn an unclear symptom into a minimal reproduction and verified root cause.

• Design a clear REST API contract

  Specify routes, validation, authorization, errors, and pagination consistently.

• Design stable API pagination

  Specify bounded pages, deterministic ordering, and clear totals for collection endpoints.

• Investigate a slow request with evidence

  Locate latency in the full request path before suggesting an optimization.

• Map repository module boundaries

  Explain ownership, entry points, dependency direction, and generated areas from the codebase.

• Map sensitive data through a feature

  Trace collection, storage, access, logging, sharing, and deletion of user data.

• Plan a repeatable, reversible data migration

  Preserve legacy rows and prove the migration can safely run again.

• Refactor while preserving contracts

  Improve a module in small steps without changing its public behavior.

• Review a change for correctness and maintainability

  Focus a code review on behavior, edge cases, and future change cost.

• Review a generated instruction file

  Check format compliance, useful context, and absence of fabricated project facts.

• Review a React flow for accessibility

  Find keyboard, focus, semantics, and screen-reader barriers in a real interaction.

• Review a schema change before rollout

  Check constraints, compatibility, migration cost, and recovery for a data model change.

• Review a security-sensitive change

  Find exploitable defects and regressions in a focused code change.

• Review a SQL query for correctness and cost

  Check query semantics, indexes, locking, and safe parameterization.

• Review AI prompt-injection boundaries

  Trace untrusted content, model instructions, tool permissions, and consequential actions.

• Review session and API-key boundaries

  Verify that credential type, CSRF, role, and ownership checks match endpoint intent.

• Review translation completeness and meaning

  Check that localized content is complete, natural, and faithful to the source.

• Turn a user need into acceptance tests

  Write observable acceptance criteria for normal, empty, invalid, and denied paths.

• Validate untrusted data at a TypeScript boundary

  Build a runtime parser that narrows unknown input and returns useful errors.