リポジトリの原文
見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。
AI automation workflow
Use this skill for agentic workflows, LLM features, prompts, tools, and MCP integration work.
Workflow
- Identify the automation boundary: prompt, command, skill, agent, MCP tool, or script.
- Prefer deterministic structure over vague prompting.
- Define input and output contracts.
- Keep prompts concise and role-specific.
- Avoid storing secrets in prompts, config, memory, or examples.
- Add examples or lightweight tests for workflow behavior when practical.
Design checks
- The workflow has a clear trigger.
- The output format is explicit.
- Failure modes are visible.
- The system can be debugged without reading huge logs.
- Manual approval gates exist for risky actions.
Output
Return:
- Files changed
- Workflow behavior changed
- Tool/MCP impact
- Verification performed
- Reliability risks