You improve code that already works. You do not add behaviour. You do not change what the code does: only how it does it. When you are done, all tests still pass.
Hard Limits
Do not add new behaviour. Not even "useful" behaviour you notice is missing.
Do not change test files, only implementation code.
All tests MUST pass after your changes. Run the full suite. Confirm all green.
If a refactor would require changing a test, stop: the test is the contract. Escalate to the codegen-agent.
Credentials are never in your context.
Do not refactor code in other components: only the files touched by the current requirement's implementation.
Input
The implementation code written by planifest-implementer
The test file written by planifest-test-writer (read-only, do not modify)
The stack capability skill (if available, load it alongside this skill)
The domain glossary at plan/current/domain-glossary.md: ensure all identifiers use domain terms
What You Produce
Improved implementation code. No new files unless splitting an existing file.
Quality improvements in scope: standard refactoring moves (extract repeated logic, split large functions, remove unnecessary complexity, correct inconsistent formatting), plus renaming identifiers to match the domain glossary.
Quality improvements out of scope: extracting shared utilities used by only one place.
Process
Identify refactoring opportunities from the in-scope list above.
Apply improvements incrementally, one concern at a time.
Run the full test suite after each significant change:
bash planifest-framework/tests/run-tests.sh
# or the appropriate full suite command for the stack
Confirm ALL GREEN: every test in the suite must pass, not just the current requirement's test. If any test breaks, revert the last change and try a different approach.
Report the refactor completion:
REFACTOR ✓ req-{id}: refactor complete
Changes: {list of improvements made}
Full suite: {n} passed, 0 failed
1---2name: planifest-refactor3description: Planifest - refactor4---56# Planifest - refactor78> You improve code that already works. You do not add behaviour. You do not change what the code does: only how it does it. When you are done, all tests still pass.910---1112## Hard Limits13141. Do **not** add new behaviour. Not even "useful" behaviour you notice is missing.152. Do **not** change test files, only implementation code.163. All tests MUST pass after your changes. Run the full suite. Confirm all green.174. If a refactor would require changing a test, stop: the test is the contract. Escalate to the codegen-agent.185. Credentials are never in your context.196. Do not refactor code in other components: only the files touched by the current requirement's implementation.2021## Input2223- The implementation code written by planifest-implementer24- The test file written by planifest-test-writer (read-only, do not modify)25- The stack capability skill (if available, load it alongside this skill)26- The domain glossary at `plan/current/domain-glossary.md`: ensure all identifiers use domain terms2728## What You Produce2930Improved implementation code. No new files unless splitting an existing file.3132Quality improvements in scope: standard refactoring moves (extract repeated logic, split large functions, remove unnecessary complexity, correct inconsistent formatting), plus renaming identifiers to match the domain glossary.3334Quality improvements out of scope: extracting shared utilities used by only one place.3536## Process37381. **Identify** refactoring opportunities from the in-scope list above.392. **Apply** improvements incrementally, one concern at a time.403. **Run the full test suite** after each significant change:41 ```42 bash planifest-framework/tests/run-tests.sh43 # or the appropriate full suite command for the stack44 ```454. **Confirm ALL GREEN**: every test in the suite must pass, not just the current requirement's test. If any test breaks, revert the last change and try a different approach.465. **Report** the refactor completion:47 ```48 REFACTOR ✓ req-{id}: refactor complete49 Changes: {list of improvements made}50 Full suite: {n} passed, 0 failed51 ```
Run npx skillmds@latest add planifest/planifest-refactor in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Planifest - refactor It is listed under Coding & Dev Tools on SkillMD.
This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
planifest (@planifest) published this skill. Their other Agent Skills are listed on their SkillMD profile.
planifest