Skip when "resume the work in <doc>" already suffices — just hand them that path.
Protocol
Step 1 — Scope the handoffs
Use the request and conversation to choose the prompt count and boundaries:
- Named items. Write one prompt per item, in order; note any sequencing.
- One thread. Write one prompt.
- Ambiguous threads. Use
AskUserQuestionto ask whether to combine or separate them, and list the proposed split.
Step 2 — Locate context to reference, don't restate
- Source. Point
Read first:at the state artifact the workflow already maintains (plan.md,interview.md, spec, or investigation); otherwise reference the source files, commit, or PR. - Delta. Add only missing current status, the last decision, and the next action. Drop completed next actions and tell the new chat to rebuild its task list from the sources.
- Missing owner. If the remaining state cannot fit a lean prompt, use
AskUserQuestionto ask whether and where to create an active-state artifact. Use a location maintained by an existing workflow or chosen by the user; never default toCONTEXT.md.
Step 3 — Draft each prompt
- Skeleton. Use only the headings below; fold needed details into an existing field.
- Fields. Put each label on its own line, place values beneath it, and bullet multiple values. Omit empty fields; include
Status:only for work in progress. - Preferences. Let the target load its own repository instructions. Carry only task-specific guardrails.
Repo:
<abs path>[ · Working dir: <dir> — never touch <x>]
Read first:
- <path>
- <path>
Status:
<one line — where things stand>
Goal:
<the one outcome this chat should reach>
Constraints:
- <read-only vs act intent — "analysis only, present for me to decide" vs "kick off the work">
- <don'ts; guardrails not guaranteed by the target's config>
Output:
<what to produce / the format expected>
Step 4 — Emit
Return each prompt in a separate copy-pasteable block; write files only when asked. Add one paste instruction and one run-order line when sequenced.