techmefr/mentis/code
Claude Code and agent skill by techmefr/mentis
- community
Checking that the SKILL.md file is still in the repository.
Use when the build step starts and the tests are written, implement in increments until they pass, one task_item at a time.
Install
npx vskill@latest install techmefr/mentis/codeSKILL.md
The same SKILL.md was published earlier as techmefr/mentis/code.
code
Step 6 of the pipeline (WORKFLOW.md). Build the minimum that makes each test pass.
When
After tdd (red tests written), during /BUILD.
Steps
- Climb the ladder before writing, after reading the task and the code it touches. Stop at the first
rung that holds: does the need exist at all; is it already in this repo; does the standard library do it; does the platform do it natively; does a dependency already installed do it; can it be one line; only then the minimum code. The ladder shortens the solution, never the reading. It never removes input validation at trust boundaries, error handling that prevents data loss, security measures or accessibility basics. The long form, with the guards, is skills/reuse-before-write.
- Take one
task_item, write the minimum code that makes its test pass. - Run that test scoped, per
skills/run-generated-tests— the new/changed test, at widest the file
it lives in, never the project's bare full-suite command. toggle_task_item when the increment is done, commit.
- Blocked / unexpected error → invoke the `debug` block before proposing a fix.
- Repeat until the
task_itemsare exhausted.
Output / checkpoint
build_done.
Guardrails
No comments in the code. Never mark a test passes: true by hand, the GATE (step 7) decides, on evidence. We don't validate ourselves.
A failing test gets the implementation fixed, never the test loosened. Editing a pre-existing test file is for adding the case this task's new behaviour actually introduces, never for deleting, loosening or retargeting an assertion that was already there to make it match what the code currently does — that specific move is how a regression ships behind a green suite (see debug §3.4/Guardrails for the full statement, and route through tdd/dozer if the test itself is genuinely the thing that's wrong).
Origin
Native Claude Code + internal, rewritten. The no-test-tampering guardrail added 2026-08-11, same change as debug/tdd and the implementer agents: named directly by the operator, not sourced from a catalogue.
The ladder (step 0) is rewritten from the ponytail skill (MIT), read 2026-10-02. Its marker comment, debt ledger and benchmark figures are not carried over: comments are forbidden here and the author's numbers are not measured on this repo.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/code/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/techmefr/mentis/code)