| name | scramble |
| description | Tactically assemble a short-term plan of many parallel options, then pursue as much local action as possible in the moment โ rebases, protogen, PR pushes, tracker state updates, sqfan fleet dispatch. Not a survey (orient/sitrep) and not a single-threaded task: a multi-option sprint where the agent fans out everything it can do locally right now. Triggers on "scramble", "max delta-v", "30 minute sprint", "parallel push", "many options at once", "as much local action as possible", "unblock and ship", "velocity push". |
scramble
Time-boxed sprint (default 30 minutes): build an option board of everything that could move, execute the largest independent set in parallel, locally, then score what changed state. Not sitrep (one goal โ A/B/C call), not orient (sweep โ gestalt), not one task run end-to-end. Bias to doing: no reports, tracker comments, or planning docs unless asked. If an architectural decision must precede code, plan first โ not a scramble.
Hard rules
- Options first โ scan everything that can move before committing to any single thread.
- Local action wins โ protogen, rebases, tests, pushes, tracker states, batch writes run on host; never wait on squire for mechanical work.
- Execute yourself โ never end with "you should run โฆ".
- Linear: state transitions only, no comments (unless overridden).
- Link PRs at every mention.
- Fan-out limit is independence โ options sharing a dirty tree or branch serialize.
Workflow
1. Pin the window (30s). Time budget (default 30m), scope, metric = count of state transitions (merged, pushed, opened, dispatched, tracker state changed).
2. Option board (2โ3 min). Tracker items in scope + real-world signals (PRs, branches, CI, blockers). One line per option: what, type, Local?, depends-on, estimate. Mark Local? honestly โ remote-blocked options stay on the board but never enter the execution set.
| Type | What it is |
|---|
| close-in-review | chain ready to land; rebase blocker, re-request review, or Done if merged |
| unblock-in-progress | branches exist; needs protogen, rebase, conflict fix, draft PR |
| fleet-dispatch | Todo item well-scoped for sqfan; write batch + fire |
| tracker-hygiene | wrong state |
| open-PR | branch pushed, no PR yet |
| local-test | targeted tests to unblock a push |
| remote-blocked | needs human review/approval/decision โ note, don't pretend |
3. Plan (1 min). Execution set: every local, independent, state-moving option; prioritize close-in-review and unblock-in-progress (highest delta-v per minute); 1โ3 fleet dispatches if they won't steal focus from local pushes; cap at 3โ6 parallel threads. Present as a numbered list (to run, not to approve), then start all independent items immediately.
4. Execute (bulk of window).
Close-in-review, per chain: map proto โ server โ SDK โ CLI PRs. All merged โ tracker Done. Blocker PR open โ rebase on main, fix conflicts, make protogen, targeted tests, push, re-request reviewers. Never Done while any chain PR is still open.
Unblock-in-progress: fetch + rebase every repo in the chain; local protogen/tests, fix only what breaks; push, open draft PRs for missing layers; tracker โ In Review once the stack is PR-visible.
c1 protogen โ always local, one at a time per clone (docker lock), so start it early alongside other repos: merge origin/main, fix .proto by hand (drop duplicate rpc stubs), make protogen, targeted go test, push.
Fleet-dispatch (Todo โ sqfan):
| Signal | Dispatch? |
|---|
| Well-scoped, โค3 repos, clear deliverable | Yes |
| Needs running backend | Yes + c1-dev-stack |
| Adjacent parity work on existing branch family | Yes |
| Decision / ADR / "decide whether" | No |
| Large platform / hardware greenfield | No |
| Blocked on in-flight stack merge | No |
| <15 min of local work | No โ do it locally |
Write sqfan-batches/<slug>/batch.yaml + prompts; dispatch fire-and-forget via sqfan MCP โ don't poll to completion during the scramble. Cap 2โ3 dispatches per scramble.
Tracker hygiene (throughout):
| Condition | State |
|---|
| Stack PR-visible, CI running | In Review |
| All chain PRs merged | Done |
| No work, wrongly In Progress | Todo |
| Decision ticket | Todo |
5. Scorecard (last 2 min).
## Scramble scorecard โ <date> (<N>m)
**Plan:** <N> options assembled, <M> executed in parallel
**Delta-v:** <one sentence>
### Executed
### Landed (state changed)
### Dispatched (fire-and-forget)
### Waiting on others (on the board, not executed)
### Not moved (honest)
Pairing
| Need | Skill |
|---|
| Which milestone is active | tactical-sitrep (before scramble if unclear) |
| c1 dispatch briefs | c1-squire-dispatch |
| sqfan mechanics | sqfan |
| CI red on blocker PR | gh-fix-ci |
Project defaults (Latchkey reviewer sets, milestone IDs, repo chains): references/latchkey.md.