/run

name: run
description: Auto-complete all pending tasks or work on specific task
allowed-tools: [Bash, Read, Glob, Grep, Write, Edit, Skill]
argument-hint: "[task_id]"

STEP 0: INVOKE SKILL (MANDATORY)

**BEFORE doing anything else, you MUST invoke the `feature-run` skill:**

Skill({ skill: "feature-run" })

The instructions below are a fallback only if the skill fails to load.

---

FALLBACK: EXECUTE TASK WORKFLOW

Start immediately. Do not ask for confirmation.

Mode Detection

**If task_id provided** (e.g., `/agent-foreman:run <task_id>`):

• Work on that specific task only
• Complete it and stop

**If no task_id** (e.g., `/agent-foreman:run`):

• Auto-complete all pending tasks in priority order
• **UNATTENDED MODE** - no questions, no stopping for errors

---

Single Task Mode

agent-foreman next <task_id>    # 1. Get task
# ... implement task ...         # 2. Write code
agent-foreman check <task_id>   # 3. Verify
agent-foreman done <task_id>    # 4. Complete

---

All Tasks Mode

agent-foreman status            # 1. Check pending count
agent-foreman next              # 2. Get next task
# ... implement task ...         # 3. Write code
agent-foreman check <id>        # 4. Verify
agent-foreman done <id>         # 5. Complete
# Loop to step 1 until done

---

Rules

| Rule | Action | |------|--------| | One at a time | Complete current before next | | No skipping | status → next → implement → check → done | | No editing criteria | Implement exactly as specified | | Never kill processes | Let commands finish naturally |

---

Exit Conditions

| Condition | Action | |-----------|--------| | All tasks processed | STOP - Show summary | | Single task completed | STOP - Task done | | User interrupts | STOP - Clean state |

Priority Order

1. `needs_review` → highest 2. `failing` → next 3. Lower `priority` number

---

UNATTENDED MODE RULES (CRITICAL)

When in ALL-TASKS mode (no task_id provided):

NO QUESTIONS ALLOWED

| Situation | Action | |-----------|--------| | Ambiguous requirement | Make reasonable decision, proceed | | Missing file/dependency | Create or skip, proceed | | Multiple options | Choose simplest, proceed | | Test failure | Note it, proceed | | Verification failure | Run `agent-foreman fail <id> -r "reason"`, continue |

**Forbidden phrases** - NEVER output:

• "Should I...?"
• "Do you want me to...?"
• "Which approach would you prefer?"

Loop Enforcement

1. **CHECK**: `agent-foreman status` - note pending count 2. **LOOP** while pending > 0:

• `agent-foreman next` → implement → `agent-foreman check <id>` → `agent-foreman done <id>`
• Pass? Continue. Fail? Run `agent-foreman fail <id> -r "reason"`, continue.

3. **EXIT**: Only when pending = 0

On Verification Failure

1. **DO NOT STOP** 2. Run: `agent-foreman fail <task_id> --reason "Brief failure description"` 3. **IMMEDIATELY** continue to next task

**CRITICAL: NEVER stop due to verification failure - use `agent-foreman fail` and continue!**