9 engineering disciplines for AI agents — TDD, debugging, code review, planning, and multi-agent coordination. Build software like it matters.

This skill is coherent and appears to do what it says: provide disciplined workflows for agent-driven development. Before installing, consider that the instructions expect access to your repository and developer tooling (git/gh, tests, local services) and will tell an agent to run commands like git push, run tests, and call local health endpoints. If you install it: 1) ensure the agent runs with least privilege (do not give it elevated access or unrestricted push permissions to critical repos), 2) prefer manual review or require human approval before any push/merge actions, and 3) verify that any automation spawned by the agent (subagents) is subject to the same safeguards. If you want stronger safety, disable autonomous invocation or require a human-in-the-loop for push/merge steps.

Type: OpenClaw Skill Name: craftwork Version: 1.0.0 The OpenClaw AgentSkills skill bundle 'craftwork' and its sub-skills are designed to enforce good software engineering practices for AI agents, including test-driven development, systematic debugging, code review, and structured planning. The instructions for the AI agent primarily involve standard development commands (e.g., `git`, `npm test`, `pytest`, `curl localhost`) and structured communication templates. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts with harmful objectives. In fact, the `requesting-code-review` skill explicitly includes a security checklist, indicating a focus on secure development practices.