agent-all-copilot

/agent-all (Copilot port)

Runs the cost-unrestricted multi-agent pipeline using Copilot CLI primitives. Phase 3 fan-out dispatches one task per wave task and awaits via Copilot's subagentStop hook (or polls list_agents).

Usage

/agent-all-copilot "add user signup form"
/agent-all-copilot docs/tasks/12-fix-login.md
/agent-all-copilot "fix flaky test" --loop --max-iter=5
/agent-all-copilot docs/tasks/x.md --no-pr --wave-size=large

Flags

Same as Claude Code: --loop, --max-iter=<N>, --max-cost=<USD>, --wave-size=small|medium|large, --no-pr, --no-brainstorm, --resume, --force, --yes, --break-condition=<spec>, --reconfigure.

When --loop is set, Phase 0 prompts the user interactively (via Copilot's ask_user) for the break-condition preset (test-auto / visual-qa / Custom shell / Composite) and offers to save the choice to .agent-all.json. Use --break-condition=<spec> to skip the prompt for one invocation, or --reconfigure to re-prompt even when a non-default value already lives in config.

Pipeline

Phase	File	Purpose
0	phases/0-preflight.md	git + .copilot/agents + config + input checks
1	phases/1-intent.md	brainstorm (chat) OR load task file; persist via store_memory
2	phases/2-plan.md	draft plan into store_memory (key=agent-all/plan) + file
3	phases/3-dispatch.md	fan out parallel task invocations per wave
4	phases/4-gate.md	dispatch reviewer tasks for spec + quality
5	phases/5-pr.md	read_bash: git branch push + gh pr create
6	phases/6-loop.md	breakCondition shell + state.iter re-entry

Rules

1. You orchestrate; phases are source of truth. Read each phase file before running it.
2. State lives in two places. .agent-all-state.json for cross-session resume; store_memory scope=repository for in-session dispatch coordination.
3. Delegate, don't reimplement. Phase 3 uses task; Phase 4 uses task; Phase 5 uses read_bash.
4. Loop is opt-in. Without --loop, Phase 6 is a no-op.
5. Hard caps: --max-iter clamped to 50; --max-cost enforced after each wave by reading Copilot's per-session cost field (if exposed via list_agents summary; else best-effort).

Copilot primitive map

Action	Copilot primitive
Read file	read_file
Write file	apply_patch
Shell	read_bash
Dispatch subagent	task
Inspect dispatched agent	read_agent, list_agents
Wait for completion	subagentStop hook OR poll list_agents
Persist plan/state	store_memory (scope=repository)
Prompt user	ask_user (where available)

On error

Same as Claude port. Additional Copilot-specific notes:

• If task tool unavailable (Copilot CLI < v0.0.380): abort with upgrade hint.
• If subagentStop hook not registered: fall back to list_agents polling every 2s.
• If store_memory quota exceeded: warn and continue using only file-based state.

When done

Print summary: phases completed, iters, cost, PR URL. Exit 0/1/2/3 per spec.

References

• references/porting-notes.md — Copilot primitive mapping rationale
• plugins/harness-floor/skills/agent-all/SKILL.md — source-of-truth pipeline