# Test

> Verify implemented behavior before review by running focused checks without changing feature scope.

- Skill: `tom-pudding/test` (Agent Skill)
- Install (CLI): `npx skillmds@latest add tom-pudding/test`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tom-pudding/test/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tom-pudding (https://skillmd.com/u/tom-pudding)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/tom-pudding/test

---


Use test mode.

Purpose:
- Validate actual behavior after implementation
- Confirm the agreed scope works as expected
- Surface failures clearly before review

Position in workflow:
- Run after `implement`
- Run before `review`

Rules:
- Do not implement features
- Do not redesign the solution
- Do not expand scope
- Prefer the smallest set of checks that verifies real behavior
- Test the implemented paths directly, not just static code structure
- If tests are missing, use existing commands, manual verification steps, or lightweight runtime checks
- If a failure suggests missing functionality, report it; do not fix it in test mode
- Only make non-functional test-related adjustments if explicitly requested in a separate implementation step
- Do not mark behavior as verified unless at least one real execution path was observed
- If testing depends on manual steps or environment limitations, state the limitation explicitly in RESULT
- Do not rewrite code, test fixtures, or app logic during test mode

What to verify:
- Primary user-facing behavior in the agreed scope
- Relevant regressions near the changed area
- Error paths or edge cases if they are part of the agreed scope

Output format:

TEST SCOPE:
- What behavior is being verified
- What is intentionally out of scope

TEST PLAN:
- Checks to run
- Why each check is sufficient

COMMANDS:
- Exact commands executed
- Manual verification steps if commands are not enough

RESULT:
- Pass/fail status for each check
- Key evidence or observed behavior
- Unknowns or limitations

NEXT STEP:
- `review` if behavior is verified
- `implement` if failures or missing behavior are found