# Code Write

> Use when you must generate predictable code that follows the patterns of an existing file (tests, config, type stubs, boilerplate). A cheap model writes it from your spec plus a reference file and saves it straight to disk, so you never read the reference or the output.

- Skill: `toolmonsters/code-write` (Agent Skill)
- Install (CLI): `npx skillmds@latest add toolmonsters/code-write`
- Raw SKILL.md: https://api.skillmd.com/api/skills/toolmonsters/code-write/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ToolMonsters (https://skillmd.com/u/toolmonsters)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/toolmonsters/code-write

---


# code-write

Run it from Bash. A spec, one or more reference files, and a target path:

```bash
~/.claude/bin/code-write --spec "Write pytest tests for UserService covering create, update and delete" --reference tests/test_order_service.py --target tests/test_user_service.py
```

Rules:
- The reference is required. Without a file to match, the output fits nothing in the project.
- Put every constraint in the spec: names, cases to cover, imports to reuse.
- Without `--target` the code is printed to stdout instead of written.
- Verify the result with a compiler, linter or test run, not by reading the whole file back.