メニュー
Validate that commands in .plan/runbook.md are executable — checks executables, config files, environment variables, and TODO placeholders
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
| name | test-runbook |
| description | Validate that commands in .plan/runbook.md are executable — checks executables, config files, environment variables, and TODO placeholders |
| allowed-tools | ["Read","Grep","Glob","Bash(which *)","Bash(command -v *)","Bash(*--version*)","Bash(test -f *)","Bash(test -d *)","Bash(printenv *)"] |
Read .plan/runbook.md. Missing → FAIL, stop. Empty → FAIL, stop.
Case-insensitive scan. These block validate_planning_exit_gate(). Record line number, section, text. Flag as blocking at top of report.
Parse fenced code blocks. For each: note section (Build/Test/Deploy/Rollback), extract command lines (skip comments/blanks), split multi-line (&&/||/|) into individual executables, note $VAR/${VAR} references.
4a. Executable: command -v <base-executable> (ignore sudo/env/nohup prefix). PASS/FAIL.
4b. Config files: Check implicit configs by tool:
npm→package.json, make→Makefile, docker build→Dockerfile, docker-compose→compose.yml, cargo→Cargo.toml, go→go.mod, pip→requirements.txt, poetry→pyproject.toml. Also check explicit paths in commands.
4c. Env vars: printenv for each $VAR. PASS (set) / WARN (not set, may be CI-injected).
4d. Version (optional): <executable> --version. PASS/WARN.
Blocking: TODO/TBD/FIXME with line numbers
| Section | Command | Executable | Status | Notes |
Env vars: variable → set/not set
Fixes: concrete remediation per FAIL/WARN
Summary: "X commands, all passed" / "Z failures, Y warnings"
NEVER execute deploy/rollback/destructive commands. NEVER run rm/delete/destroy/drop/kill. NEVER pipe to sh/bash/eval. Only: which, command -v, --version, test -f, printenv.
Show planning area readiness status -- statuses, open questions, blockers, and next actions
Clone an existing Evoloop story as a template for a new one — copies structure, resets status, prompts for changes to title, requirements, and dependencies
Estimate the total prompt size that will be sent to an AI agent when processing an Evoloop story — warns if approaching model context limits
Visualize the Evoloop story dependency graph as text tree and Mermaid flowchart — shows critical path, blocked chains, and ready-to-run stories
Run Evoloop doctor checks and auto-fix common issues — missing directories, non-executable scripts, missing templates, and provide remediation for manual fixes
Deep-dive into a specific story's state — stage, failures, dependencies, last agent output, requirement verification, and suggested next action