メニュー
Conventions for disposable scratch code and verification scripts. Use when writing smoke tests, quick probes, or temporary artifacts during task execution.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Multi-model supervisor that discovers skills, picks models, and composes runs. Use when executing multi-step plans across multiple models.
Agent execution engine that composes prompts, routes models, and writes run artifacts. Use when launching subagent runs.
Multi-model supervisor that discovers skills, picks models, and composes runs. Use when executing multi-step plans across multiple models.
Agent execution engine that composes prompts, routes models, and writes run artifacts. Use when launching subagent runs.
Verifies implementation aligns with stated requirements and acceptance criteria. Use before implementation (to clarify scope) and before final sign-off (to detect gaps).
Breaks the next task from a plan into an implementable task file. Use when decomposing a multi-step plan into ordered work units.
| name | scratchpad |
| description | Conventions for disposable scratch code and verification scripts. Use when writing smoke tests, quick probes, or temporary artifacts during task execution. |
| user-invocable | false |
Disposable code and artifacts live in .scratch/, either in the working directory or under a designated output directory. Never committed.
Any disposable code used to quickly verify something works. Write them liberally after implementing changes.
| Type | When to use | Example |
|---|---|---|
| Logic verification | Verify guard patterns, state machines, edge cases | vitest test asserting a staleness guard rejects stale responses |
| API/endpoint probes | Verify endpoints respond correctly | curl scripts hitting local dev server |
| Integration checks | Verify two systems work together | Script that creates a resource then reads it back |
| Regression probes | Verify a specific bug is fixed | Minimal repro of the bug, now passing |
.scratch/ — never commit it.env values, JWTs, API keys, cookies)CLAUDE.md or AGENTS.md) for auth token scripts and API base URL{feature}-{what}.test.ts or {endpoint}.shWrite markdown notes as you work — findings, decisions, progress. Notes persist on disk across context clears.
Use the co-located script:
# Write a quick note (auto-named with timestamp)
scripts/scratch.sh write --tag research -p "Found that X uses Y pattern"
# Write from stdin
echo "curl returned 200" | scripts/scratch.sh write --tag probe
# Explicit file path within session
scripts/scratch.sh write my-test.sh -p "#!/bin/bash\necho test"
# List notes from current session
scripts/scratch.sh list
# Read latest note
scripts/scratch.sh read @latest
# List all sessions
scripts/scratch.sh list --all
When run via run-agent, session defaults to the run ID. Standalone, pass --session or let it auto-bucket by date.