菜单
Investigate and engage concurrently: don't block on evidence before surfacing your read.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | explore-and-engage |
| type | principle |
| description | Investigate and engage concurrently: don't block on evidence before surfacing your read. |
| user-invocable | true |
<explore_then_recommend> Ground recommendations in evidence from all sources: code, docs, KB, prior decisions, and the running system. Docs and code go stale: probe real behavior when it matters. When you lack evidence, investigate before forming a view. </explore_then_recommend>
In primary mode (interactive with the user), run two tracks concurrently:
The tracks feed each other. Exploration findings reshape your questions. User answers redirect your investigation. Neither waits for the other.
In subagent mode, the Explore discipline still applies: investigate before concluding. The Engage track doesn't apply; there is no human in the loop.
Use when grilling should be grounded primarily in project artifacts. Loads `/grill`, checks project vocabulary and prior decisions first, and spawns bounded explorers when the answer should come from docs.
Load for large human-facing outputs: reports, docs, multi-section explanations, artifacts. Choose the right form for each beat (prose, diagram, table, mockup) and put the answer first, depth behind it. If the medium allows, use progressive disclosure.
Use when deciding where knowledge goes or reading/writing durable docs: AGENTS.md, .context/, KB, docs/, and work directories.
Anchor work in a work item: artifacts on disk, not in context.
Use when exploring or changing the codebase: read AGENTS.md first, use .context/CONTEXT.md for detail, keep intent docs succinct.
Use when recovering decisions or context from conversation history: session log navigation, transcript delegation.
