Quality

/large-plan

Invoke before any git/gh and before starting another agent (delegate_task, new thread, sub-agent): make a plan, plan next steps, lead kickoff, feature batch, multi-PR; dispatch a worker, send a brief, fan out.

$ golems-cli skills install large-plan
20 evals
fixtures
5 workflows

Updated today

Invoke as: /large-plan (single segment).
Source: $HOME/.golems/skills/golem-powers/large-plan/ (symlinked at ~/.claude/skills/large-plan).

Fleet law: canon #2 owns PR-loop validity and canon #7/#8 own collab/lead-routing law. This skill keeps plan scaffolding, phase routing, collab file mechanics, adapters, and evaluator gates.

Scaffold folder-based plans with phase folders, execute them through the branch-PR-review cycle, and coordinate async agent collaboration.

Kickoff contract

  • Lead launch instructions live in /agent-routing § Service tier and placement: "Read and follow /LEAD.md and the global LEAD.md" (~/.codex/LEAD.md).
  • Claude seats: invoke /large-plan with the Skill tool before the first plan action of a lane: making a plan, planning a feature batch or next steps, or kicking off a lead. Make it the first tool call, before any git/gh/grep orientation: live misses open with git status/gh pr and never load the skill (PR15 trigger eval, 2026-10-19).
  • Codex seats: read this SKILL.md body (a file read) before that action; Codex has no Skill tool.
  • A lead's kickoff brief to a worker names the skills the worker invokes (/large-plan, /agent-routing, /pr-loop).
  • Standing ask: "Once its PRs land, invoke the updated skills at the start of each plan, dispatch or PR action." (Etan 2026-10-18, via orc collab)
  • Why: Claude T3 leads made 0 Skill calls on Oct 16–18 while all three lead skills sat in their catalog (skill-edits audit, 2026-10-18).

Before every dispatch

Check this at each dispatch, not only at scaffold time: a lane added later counts the same as a planned round. When a worker, lane or sub-agent is already running, or this turn dispatches 2+:

  1. collab.md exists at the plan root. If not, write it now from workflows/collab.md; a short plan's collab is a few lines.
  2. This brief contains Report to <collab path>, not to the lead thread.
  3. STATUS.md lists the new lane. Done = all three are true before the delegate_task / t3_thread_launch / Agent call. Otherwise, don't dispatch yet.

Every worker, always (a first single worker too): the dispatched thread's title starts with the seat prefix (e.g. skillcreator-w93), and the brief's LAST step is on the worker's own thread: t3_thread_organize settle (its own threadId) if its work is merged or handed back, or snooze to a stated time if its PR is still open. Its DONE line says which (Thread: settled or Thread: snoozed until <ISO>).

At every turn end (lead)

Before ending a turn, on EACH Mac (environmentId for the other one): t3_thread_list {allProjects: true, settled: false, snoozed: false, includeSubagents: true, titleContains: "<seat prefix>"}, following nextCursor until it is null. Delegated children count as subagents, so includeSubagents: false hides exactly the threads Etan sees idle; the prefix is why every dispatched title carries it (an older owned thread without it: settle or retitle it once). Every listed thread that is not running gets dealt with now, through the /agent-routing § T3-native dispatch #4 sequence (verify and harvest its docs.local, task_cancel, unwatch_pull_request, then settle with its explicit threadId), or is snoozed to when it is next needed, with the reason in the plan. A thread Etan snoozed himself is left alone (#7 there). Scheduled-task threads snooze themselves to their next run (#10 there). (Etan, 2026-11-13, via orc: "So many woke threads"; a promise did not hold, so this is a check, not a promise.)

Workflows

/large-plan:collab/large-plan:execute-phase/large-plan:phase-evaluator/large-plan:scaffold/large-plan:ux-ui