القائمة
After team plan injection, dispatch one full-width delegate_task batch immediately, notify the user how to watch the swarm, and avoid wasted coordinator turns before synthesis.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Keep coordinator and leaf context lean-batch reads, cap logs, use execute_code to filter-so swarms stay fast and within token limits.
Use at the start of non-trivial software tasks. Gather repo ground truth with batched reads and searches before planning or delegate_task.
Obey Plan 1A hook session phases so mutating tools, delegate_task sizing, and re-planning do not waste turns or trigger blocks.
Use when spawning delegate_task leaves for coding work. Pack self-contained goals and context so parallel Composer workers ship verifiable patches without coordinator chat history.
Require structured candidate-result JSON from swarm leaves so critique_candidates.py and synthesis run on evidence, not prose self-reports.
Enrich slim hook-injected delegate_tasks with touch map, failure excerpts, and verification commands so parallel leaves produce verifiable patches without coordinator chat history.
| name | heavy-swarm-dispatch |
| description | After team plan injection, dispatch one full-width delegate_task batch immediately, notify the user how to watch the swarm, and avoid wasted coordinator turns before synthesis. |
| version | 0.1.0 |
| author | CodeGraphTheory |
| license | MIT |
| metadata | {"hermes":{"tags":["heavy-coder","delegation","performance","coordinator"],"related_skills":["heavy-team-default","heavy-hook-phases","heavy-pre-dispatch-enrich","heavy-leaf-brief"]}} |
Hooks inject DELEGATE_TASKS_JSON and set phase AWAITING_DELEGATE. The coordinator's job is one parallel delegate_task call with the entire delegate_tasks array-then get out of the way until the batch completes.
pre_llm_call injection or manual team_coordinator.py outputAWAITING_DELEGATE and you have not yet dispatchedheavy-scope-router)Do not use when:
heavy-scope-routerAWAITING_SYNTHESIS or complete)Confirm width - Match plan width and profile minimum (heavy_coder.council_width, min_delegate_tasks; default 8). Never shrink the batch to "save cost" unless user chose single mode.
Optional enrich - If hook context is slim, run heavy-pre-dispatch-enrich before this step (same turn if possible).
Dispatch once:
delegate_task(tasks=[ ... all entries from plan delegate_tasks ... ])
Pass goal, context, and toolsets from each plan entry. Do not spawn leaves one-by-one.
User notice (same response as dispatch, before end of turn):
/agents for live subagent dashboardpython scripts/swarm_watch.py --repo .While waiting - Coordinator may: answer user, read files for synthesis prep, update todo. Do not: patch, write_file, write-like terminal, re-run full team_coordinator.py for the same unchanged task.
On "ASYNC DELEGATION BATCH COMPLETE" - Load heavy-synthesize-winner; do not start a new council for the same task.
Done when: delegate_task count ≥ required minimum and user knows how to observe progress.
| Symptom | Fix |
|---|---|
| "I'll delegate in the next turn" | Dispatch now in the same turn as the plan |
Partial tasks=[c1,c2] only | Full array from plan |
| Re-planning after injection | Trust plan file under .heavy-coder/plans/ |
| Silent background swarm | Always give /agents + watch hint |
pre_tool_heavy_team.py blocks undersized delegate_task and repo writes before delegation. Undersized batches waste a blocked turn-size the array correctly the first time.