Run Execute is an operational skill for driving spec-driven work runs: it initializes runs, enforces scripts for init/complete, generates implementation plans, runs tests, invokes code review, and finalizes artifacts. It supports single and batch runs with clear artifact timing (run.md, plan.md, test-report.md, review-report.md, walkthrough.md).
Use this skill when you need reproducible, auditable execution of development work items in CI-like flows: autopilot runs for quick execution, confirm mode for human-approved plans, and validate mode for design-driven implementations. It’s designed for engineering workflows that require strict checkpoints and artifacts.
Best for agent runtimes that can run repository scripts (Node.js scripts) and manage filesystem artifacts; complements code-review and walkthrough-generation skills.
Run-execute orchestrates work item execution across autopilot/confirm/validate modes with init/complete scripts, plan/test/report artifacts, and strict gating. Scripts are well-written Node.js CommonJS with thorough validation and error handling, but couldn't be executed because the auditor's run_scripts.py doesn't support .cjs extension. SKILL.md is detailed and prescriptive with clear XML-tagged sections, hard gates, and artifact timing rules.
yaml (npm package, required by all scripts)Scripts do filesystem I/O (read/write state.yaml, create run folders) but validate paths and handle errors gracefully. No network calls, no credential handling, no destructive operations. The .cjs extension issue is a tooling gap in the auditor, not a skill defect. Skill is well-suited for its niche (FIRE/specs.md workflow) but has narrow applicability outside that ecosystem.
