Codex & Claude Code prompt library

Ready-to-copy prompts for Codex and Claude Code, grouped by what they're for. Each one includes scope limits and safety constraints (new branch, run tests, no push) so it's safe to hand off and walk away from.

Testing

Critical-Path Integration Test Sweep

Have the agent map the app's main user-facing flow, find where it has no integration test coverage, and add tests for the highest-risk gaps only.

Tools: Both · Updated 2026-09-22
Edge-Case Test Gap Finder

Have the agent read a module closely and list the edge cases (empty input, boundary values, concurrent access, malformed data) it can currently break, for you to confirm before any test gets written.

Tools: Claude Code · Updated 2026-09-22
Flaky Test Root-Cause Report

Have the agent run the suite enough times to catch intermittent failures, diagnose why each one flakes, and hand back a report — no fixes applied until you've reviewed the causes.

Tools: Both · Updated 2026-09-22

Cleanup

Dead Code Removal Plan

Have the agent locate unreachable functions, unused exports, and orphaned files, then produce a removal plan with a confidence rating per item — nothing gets deleted until you say so.

Tools: Both · Updated 2026-09-22
Duplicate Logic Consolidation Report

Have the agent locate blocks of near-identical logic copy-pasted across the codebase and propose a single shared home for each, without merging anything yet.

Tools: Codex · Updated 2026-09-22

Docs

Inline Comment Audit for Non-Obvious Code

Have the agent find code whose behavior isn't self-explanatory and add comments explaining the reasoning behind it, while explicitly avoiding comments that just restate what the code already says.

Tools: Claude Code · Updated 2026-09-22
README Onboarding Rewrite

Have the agent actually try to follow the current README's setup steps as written, note where it breaks or assumes tribal knowledge, and rewrite it into a quickstart that works.

Tools: Both · Updated 2026-09-22

Audit

Dependency Upgrade Risk Audit

Have the agent check which dependencies are behind, read their changelogs for breaking changes, and rank upgrade risk — report only, no version bumps applied.

Tools: Both · Updated 2026-09-22
Error Handling & Logging Consistency Audit

Have the agent survey how errors are caught, logged, and surfaced across a module or service, and flag where the approach is inconsistent — report only, no changes applied.

Tools: Both · Updated 2026-09-22
Public API Surface Naming Audit

Have the agent list every publicly exported function, class, and type in a module, then flag names that are too generic, inconsistent with sibling names, or hard to find by search — report only.

Tools: Codex · Updated 2026-09-22

Refactor

Config Sprawl Consolidation Plan

Have the agent locate configuration values and constants that are duplicated or hardcoded in multiple places, and propose a single consolidated source — plan only, no changes applied.

Tools: Claude Code · Updated 2026-09-22
Large Function Decomposition Plan

Have the agent break down what an overgrown function is actually doing, propose a split into named pieces, and wait for your sign-off before touching any code.

Tools: Both · Updated 2026-09-22

Tooling

One-Off Scripts to a Single CLI Tool

Have the agent gather a pile of scattered one-off scripts into one CLI tool, one subcommand per original script, with argument parsing and --help instead of hand-edited constants.

Tools: Codex · Updated 2026-09-22
Pre-Commit Hook Setup for an Existing Repo

Have the agent look at what lint, format, and test tooling a repo already uses, and propose a minimal pre-commit hook setup built from those — nothing installed until you approve it.

Tools: Both · Updated 2026-09-22