Rigor Queue Survey

Rigor queue — survey coverage

Orchestrator-led, multi-turn, in-session queue. Primary entry: user runs pi in the Rigor repo (packages in .pi/settings.json) then /queue-survey or /skill:rigor-queue-survey. Same project session continues with pi -c.

Not a one-shot architect→lane.

See also: ../../roles/orchestrator.md, ../../contracts/README.md (QueueTurn), project agents under .pi/agents/ (rigor-lane, rigor-reviewer).

Requires project package npm:pi-subagents (listed in .pi/settings.json).

Hard rules (always)

• Survey targets under measurement need disjoint checkouts; never share a measuring target across agents. Encode this in every spawn / LaneInput.
• Managed worktree of rigor ≠ survey exclusivity. A pi-subagents managed worktree isolates the rigor implementation checkout only. It does not satisfy exclusivity of survey measurement trees such as ~/repo/ruby/rigor-survey/<project> (or /Users/megurine/repo/ruby/rigor-survey/<project>). Each measuring agent still needs its own disjoint survey target checkout; put that path in every LaneInput / spawn task.
• Default survey root: ~/repo/ruby/rigor-survey (or /Users/megurine/repo/ruby/rigor-survey). Ask if unclear; accept a path from the user.
• Prefer filing or linking Issues for holes (ADR-98), then sequential着手.
• Lanes: push head SHA and stop; this parent owns CI. No parallel host make verify.
• GitHub API: sparse CI polls (gh pr view --json statusCheckRollup), ≤1/min/PR.
• Managed worktree fanout needs a clean source rigor checkout. Do not auto-merge worktree patches without a human.

Turn protocol (in this same session)

Every turn:

1. Restate queue goal + constraints (survey root; disjoint survey target rule — separate from rigor managed worktrees).

2. Show ranked backlog of coverage holes with evidence (survey paths, gh issue links). Prefer Issues over ad-hoc path lists when possible.

3. Propose exactly one next unit of work (unless parallel spawn requested).

4. Wait for user: next / do #N / skip / stop (JA: 次 / やる #N / スキップ / 止めて / ストップ). Also accept spawn / spawn N / 全部やれ / parallel.

5. On go — prefer pi-subagents spawn (include disjoint survey target path in every LaneInput). Managed worktree: true applies to the rigor repo only; still assign unique survey checkouts.

   Caveat: use workflowScript + runs.run / runs.all for worktree: true (documented path). Direct { agent, task, worktree: true } is not the documented isolation path.

   Single:

   subagent({
     async: true,
     worktree: true,
     workflowScript: `
       return runs.run("lane", {
         agent: "rigor-lane",
         task: args.task,
         worktree: true
       })
     `,
     args: { task: "<LaneInput including disjoint survey target path>" }
   })

   Parallel (spawn N / 全部やれ / parallel):

   subagent({
     async: true,
     worktree: true,
     workflowScript: `
       return runs.all([
         { key: "h1", agent: "rigor-lane", task: args.t1, worktree: true },
         { key: "h2", agent: "rigor-lane", task: args.t2, worktree: true }
       ])
     `,
     args: { t1: "<LaneInput + survey checkout A>", t2: "<LaneInput + survey checkout B>" }
   })

   Fallback if pi-subagents unavailable: LaneInput + ./agents/pi-harness/scripts/run-role.sh lane with disjoint targets.

6. After children return: collect head SHAs / handoff manifests; parent owns sparse CI polls; children must not CI-watch. Refresh queue, print Queue: N remaining | Next candidate: … | say next/stop.

7. Stay in this session — no wrapper restart. Resume with pi -c then /queue-survey if needed.