菜单
Checkpoint the current conversation back to its Swarm channel. Use ONLY when the user explicitly asks to checkpoint, share, or save their work to Swarm — never invoke automatically, because sharing is gated on the channel's capture mode and the user's intent.
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 | swarm-checkpoint |
| description | Checkpoint the current conversation back to its Swarm channel. Use ONLY when the user explicitly asks to checkpoint, share, or save their work to Swarm — never invoke automatically, because sharing is gated on the channel's capture mode and the user's intent. |
| allow_implicit_invocation | false |
Checkpoint the current conversation to its Swarm channel, following the guidance field returned by get_context_for_injection:
share_session with trigger=agent_auto.share_session with trigger=user_request.You hold this conversation; the server never sees the transcript, so what you write IS the record. Author the content as a faithful, complete account of what actually happened — the decisions and their rationale, what changed, what was learned — not a one-line stub. A thin capture produces thin (or empty) record suggestions, so the value you bank is the value you write. summary is the injection-ready one-liner; content is the full account.
A checkpoint preserves the conversation; the durable knowledge inside it should also become records. After sharing (and even when the mode is Off), scan what this conversation produced for record-worthy canon:
If a conversation cleared that bar, name the specific record(s) and offer to promote them with promote_context_to_record (it keeps provenance back to this conversation). Recognizing canon is proactive; creating it is not — offer good candidates and let the user decide. Skip the offer when the capture mode is Off, or when no human is present.