Skip to main content

delegate

Autonomous execution — classify, allocate, and execute without confirmation

Quellinformationen

Repository
rlaope/bestwork-agent
Letzte Quellaktivität
3. April 2026 um 01:15
Erkannte Sprache von SKILL.md
Englisch
Sterne
1
Forks
0

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
delegate
description
Autonomous execution — classify, allocate, and execute without confirmation
When this skill is invoked, you MUST follow this exact output sequence. Do NOT ask for confirmation at any point. ## Step 1: Header (print IMMEDIATELY) ``` [BW] delegate engaged — executing autonomously... ``` ## Step 2: Classify and allocate (silent, no user prompt) Analyze the user's task: 1. Split into sub-tasks (1-5) 2. Detect domains per sub-task 3. Assign agents per sub-task (tech only — no PM/critic overhead in delegate) 4. Pick the fastest execution path Print the plan inline (do NOT wait for confirmation): ``` [BW] delegate plan: 1. "{task}" → bestwork:{agent} 2. "{task}" → bestwork:{agent} [BW] executing... ``` ## Step 3: Execute immediately For single task: - Load the specialist's prompt from `prompts/{role}/{name}.md` - Execute directly as that specialist For multiple tasks: - Spawn all agents in parallel using `run_in_background: true` - For EACH agent, call TaskCreate for live progress spinner: `TaskCreate: subject="bestwork:{agent} (task N)", activeForm="bestwork:{agent} working..."` - Print each launch: ``` [BW] ▶ delegate: bestwork:{agent} (task 1) [BW] ▶ delegate: bestwork:{agent} (task 2) ``` ## Step 4: Results (print as each completes, TaskUpdate to completed) ``` [BW] ✓ bestwork:{agent} done (task 1) — {summary} [BW] ✓ bestwork:{agent} done (task 2) — {summary} ``` ## Step 5: Self-verify (quick sanity check, no critic agent) After all tasks complete, do a quick self-check: 1. If code was written: run `npx tsc --noEmit` or equivalent type check 2. If tests exist: run the project's test command 3. If files were created: verify they are importable Print result: ``` [BW] delegate verify: {PASS|FAIL} ``` If FAIL: fix the issue immediately (1 retry max), then move on. ## Step 6: Summary ``` [BW] ═══════════════════════════════════ [BW] delegate complete: {N} tasks, {M} agents [BW] ═══════════════════════════════════ ``` ## Rules - NEVER ask for confirmation — that is the entire point of delegate - NEVER spawn PM or critic agents — delegate is speed-optimized - Tech agents only — pick the best-fit specialist per task - Self-verify replaces critic review (type check + test run) - Max 1 self-fix retry on verify failure - Do NOT write meeting logs — delegate skips ceremony - If the task is ambiguous, make a reasonable choice and proceed (do not ask)
Auf GitHub ansehen