SKILL.md
产品上线雷达工作流 Codex Adapter
This skill ports the original managed-agent cookbook into Codex workflow form.
Source Files
- Cookbook root:
managed-agent-cookbooks/launch-radar - Read first:
managed-agent-cookbooks/launch-radar/README.md - Agent blueprint:
managed-agent-cookbooks/launch-radar/agent.yaml - Steering examples:
managed-agent-cookbooks/launch-radar/steering-examples.json - Subagent blueprints:
managed-agent-cookbooks/launch-radar/subagents
How To Use
- Read the cookbook
README.mdandagent.yamlbefore running the workflow. - Translate Claude managed-agent concepts into Codex execution:
- Use local file reads/writes for repositories, trackers, tables, and source documents. - Use Codex subagents only if the user explicitly asks for parallel agents or the workflow itself is requested as a multi-agent run. - Use Codex automations only when the user asks to monitor, remind, watch, or repeat the workflow over time.
- If the cookbook references subagents, read only the relevant YAML files under
subagents/. - Preserve legal review limits: all outputs are lawyer-review drafts; verify current legal facts and deadlines before reliance.
- Produce the cookbook's intended artifact, such as a grid, tracker update, alert, digest, or memo, in the user's requested location or inline if no location is specified.