resume-handoff

§ 01 · purpose

Purpose

The other half of the create-handoff pair. Reads a handoff document plus every artifact it points at, re-validates that the working tree still matches the handoff's claims, and proposes the next action. The new session then re-enters with full grounded context, not just memory.

§ 02 · when to use

When to use it

• You're starting a fresh session and need to continue from yesterday's work.
• Someone else handed you a handoff doc and you need to pick up cold.
• Skip when there's no handoff. Start with discover / research instead.

§ 03 · inputs

Inputs

handoff path

Plans, research, and solutions linked from the handoff are read directly.

§ 04 · outputs

Outputs

§ 05 · key steps

Key steps

1. 01 Read the handoff and every linked artifact fully Why: Plans, research, and solutions referenced by the handoff are read directly (no skill dispatch) so the resume agent starts with the same evidence base the previous session had.
2. 02 Spawn focused research agents to refresh artifact context Why: Parallel general-purpose agents re-read artifacts and extract decisions concurrently, so context warm-up is fast even with multi-document handoffs.
3. 03 Verify working-tree state against "Recent changes" / "Learnings" Why: Files cited by the handoff are re-read at HEAD; git log/git diff is used to detect drift since the handoff was written. Surfaces deltas before the agent commits to "continue".
4. 04 Synthesize and present the current situation Why: A side-by-side of handoff status vs. current state gives the developer a chance to redirect before the resume agent acts on stale assumptions.
5. 05 Confirm next action, then resume Why: Action items from the handoff aren't trusted blindly. The resume agent re-proposes the next step against verified state and waits for confirmation.

§ 06 · related skills

Related skills