القائمة
After failed candidates or a red ship gate, spawn a narrow focused delegate_task wave with failure evidence instead of re-running a full-width council.
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-repair-wave |
| description | After failed candidates or a red ship gate, spawn a narrow focused delegate_task wave with failure evidence instead of re-running a full-width council. |
| version | 0.1.0 |
| author | CodeGraphTheory |
| license | MIT |
| metadata | {"hermes":{"tags":["heavy-coder","delegation","debugging","performance"],"related_skills":["heavy-synthesize-winner","heavy-ship-gate","heavy-leaf-brief","systematic-debugging"]}} |
Re-running eight leaves on the same failing test burns tokens without new information. A repair wave is 1-3 focused delegate_task goals with the failing command output, current diff, and a tight hypothesis-then ship gate again.
heavy-synthesize-winner produced a merge but heavy-ship-gate exit code ≠ 0Do not use when:
heavy-explore-first + systematic-debugging firstReproduce on coordinator tree - Run the failing command; capture exit code and ≤80 lines (heavy-context-budget).
Freeze scope - One bug, one test file, or one API contract. State what is out of scope.
Design 1-3 goals (distinct angles, not full role rotation):
| Slot | Angle |
|---|---|
| A | Minimal fix for failing assertion only |
| B | Test-first: adjust test if spec was wrong |
| C | Adjacent call path / regression sibling |
Context packet - Include: repro command output, git diff --stat, touch map, evidence bar (single command), link to prior winner candidate_id if any.
Dispatch - delegate_task(tasks=[...]) with count ≤ 3 unless user requests wider repair.
Note: Hooks may still enforce minimum width from profile. If blocked, user may say single mode for coordinator-only fix, or coordinator fixes with evidence after explicit single-mode opt-in.
Synthesize - heavy-synthesize-winner on repair results only; do not merge unrelated prior candidates.
Ship - heavy-ship-gate must pass before "done".
Done when: failing command passes on coordinator tree or blocker is reported honestly.
Escalate to full council (8+) only when:
| Symptom | Fix |
|---|---|
| Same failure after repair | Deeper systematic-debugging; read definition sites |
| Full swarm again | Repair wave first |
| Ship without re-run | Ship gate on coordinator tree |