/wrap — Session Wrap Skill

Invocation: /wrap Works for: any agent (saga, freya, mimir, idun)

Purpose

Saves the current session state to a file in the project repo so the next session can resume exactly where this one left off — no re-scanning, no re-discovering the obvious.

State is stored locally by default and ignored by git. The user can opt in to sharing it via git with one command.

Behavior When Invoked

1. Detect Active Agent

Identify which agent is currently active from the agent's own persona and identity. The state file will be named after that agent (e.g., _bmad/_state/saga.md, _bmad/_state/freya.md).

2. Conversational Wrap

Ask the following four questions one at a time, waiting for the answer before asking the next. Keep it conversational — this is a closing conversation, not a form.

Question 1:

What were you working on?

(Prompt if needed: which phase, which artifact, where in the process)

Question 2:

Any open questions or blockers?

(Prompt if needed: things that need answers before we can move forward, decisions that are pending, external dependencies)

Question 3:

What's the very first thing to do next session?

(This needs to be specific enough to act on immediately — not "continue the brief", but "finish the competitive landscape section and draft the positioning statement")

Question 4:

Any context I'd lose if we started fresh?

(Prompt if needed: key decisions made, constraints discovered, things the user explained that aren't written down anywhere, approaches ruled out and why)

3. Write State File (Primary)

Write _bmad/_state/[agent].md in the current project repo. Overwrite if it already exists.

Use this exact format:

# [Agent] — Session State
**Repo:** [repo-folder-name]
**Wrapped:** [YYYY-MM-DD]

## Where I Left Off
[Answer from Question 1]

## Open Questions
[Answer from Question 2 — or "None" if no blockers]

## Next Action
[Answer from Question 3 — specific enough to act on immediately]

## Context
[Answer from Question 4 — key decisions, constraints, anything needed to not re-discover the obvious]

4. Push to Agent Space (if available)

If Agent Space is configured, do two things in this order. If any step fails, log the error and continue — local state file is already saved.

4a. Push design-process folder

Scan the project for a design-process/ folder (or equivalent WDS artifact folder). Push all .md files to Agent Space via repo-files:

curl -X POST "{BASE_URL}/functions/v1/repo-files" \
  -H "Authorization: Bearer {API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "action": "put-batch",
    "project": "[repo-folder-name]",
    "files": [
      { "path": "design-process/A-Product-Brief/product-brief.md", "content": "..." },
      ...
    ]
  }'

4b. Update presence with session state

action: update-status
agent_id: [agent]
repo: [repo-folder-name]
last_status_report: [full state file content]
working_on: [Next Action — one line]
status: offline

5. Confirm

Print:

State saved to `_bmad/_state/[agent].md`

To preserve across sessions: git add -f _bmad/_state/[agent].md && git push

Notes

The _bmad/_state/ folder is gitignored by default. The user must explicitly opt in with git add -f to share state across machines or with collaborators.
Local state file is always written first — Agent Space sync is optional and non-blocking.
If the agent can't determine the current project repo, ask the user to confirm which project they're in before writing.
Do not add a summary or closing remarks after the confirmation — the wrap is complete.

3.7 KiB Raw Blame History