harness-plan

name	harness-plan
description	Agent-native harness plans — lakes/context bundles, planning context, parallel_probes debate profile, plan-synthesizer on low/med risk, path-first approve_plan/create_plan, then DAG + debate.

harness-plan

Use this skill's phase order, spawn laws, and artifact contract directly.

When to use

• /harness-plan, harness-auto plan phase, drift replan, policy-gate without approved plan

Team topology (spawn laws)

1. Parallelism law — Parallel tasks only for independent lanes (implementation ∥ stack ≤2). Never parallelize debate or decompose ∥ hypothesis.
2. Two-pizza cap — Max 1 debate agent, 1 optional planning-context subagent, per subagent call.
3. No redundant thinkers — Read upstream YAML; do not re-run graphify in decompose when planning-context architecture coverage is ok.
4. Sequential chain — task clarification → planning context → decompose → hypothesis → research → author → DAG → debate → approve.
5. Tool intelligence — Parent picks graphify, sg, ccc by task; no mandatory tool-tied scout subprocesses.

Workflow (parent orchestrator)

1. Phase 0: artifacts/task-clarification.yaml — investigate (code + web OK), ask_user until unambiguous, gate before any planning subagent.
2. Phase 1: Compile artifacts/planning-context.yaml with tools (default) or optional planning-context subagent; inherit Phase 0 grounding.
3. Sequential decompose → gate artifacts/decomposition.yaml.
4. Sequential hypothesis (requires decomposition).
5. Phase 3.5: implementation-research.yaml + stack.yaml (parent inline and/or parallel researchers).
6. Draft PlanPacket shell; ask_user on material fork after Phase 3.5 (research-backed; not Phase 0).
7. execution-plan-author → merge execution_plan.
8. validate-plan-dag.mjs (must pass).
9. harness_plan_debate_eligibility — parallel_probes spawns plan-evaluator ∥ plan-adversary, then integrator round.
10. approve_plan({ human_summary? }) / create_plan() — packet from plan_packet_path on disk (path-first).

--quick skips semantic coverage in planning context and post-run adversary only — not adequate reconnaissance, implementation/stack artifacts (med/high risk), or plan debate.

Rules

• On-disk plan artifacts are YAML (task-clarification.yaml, plan-packet.yaml, research-brief.yaml, planning-context.yaml).
• Phase 0 allows codebase + web reads; blocks planning subagents and plan artifacts until clarification is ready.
• Subagents read-only; parent writes run artifacts and calls approve_plan / create_plan.
• context-mode only on harness paths.
• Phase 3.5 artifacts required for med/high risk unless documented waiver.

Output

plan_status, plan_review_path, next_command: /harness-run when ready.