--- name: handoff description: Compact the current conversation into a handoff document for another agent to pick up. Use when you need a recap, a directed attended handoff, or a `/goal` prompt from an existing execution spec. argument-hint: What will the next session be used for? "[recap | directive | goal [focus]]" --- # handoff If the user gives a file path, read that file as the conversation source instead of the current conversation. | Argument | Mode | Reference | |---|---|---| | empty or `recap` | Recap | `references/recap.md` | | `directive ` or bare text | Directive | `references/directive.md` | | `goal [focus]` | Goal | `references/goal.md` | ## Recap and directive Write a handoff document summarising the current conversation so a fresh agent can continue the work. Save it to `~/.jstack/handoffs/session-YYYY-MM-DD-.md` (add `-2`, `-3`, and so on when needed). Include a `## Suggested skills` section in the document. Every recap or directive handoff document must include this receiver instruction near the top, before the task-specific instructions: ```text Read and restate your understanding before proceeding. ``` Replace `` with the absolute path of the handoff file that was actually saved. Expand `~`; do not leave the placeholder or a relative path in the generated document. The receiving session must restate its understanding before it performs investigation, writes, or external actions. Reference content already captured in specs, plans, ADRs, issues, commits, or diffs by path or URL. Redact sensitive information, such as API keys, passwords, and personally identifiable information. Treat the directive or bare argument as the next session's focus and tailor the document to it. ## Goal Use goal only when an execution spec already exists. If it does not, direct the user to `/to-spec`. Goal returns the prompt described in its reference; the spec remains the source of truth. ## Next skills | Next | When | |---|---| | `to-spec` | The receiving work needs an execution contract. | | `querybrain` | The receiving work needs second-brain context. |