team-shinchan:autopilot
npx machina-cli add skill seokan-jeong/team-shinchan/autopilot --openclawFiles (1)
SKILL.md
2.6 KB
EXECUTE IMMEDIATELY
Output immediately before executing:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
👦 [Shinnosuke] Autopilot mode engaged~ 🤖
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Step 1: Validate Input
If args length > 2000 characters:
Truncate to 2000 characters
Warn user: "Request was truncated to 2000 characters"
Step 2: Execute Task
Do not read further. Execute this Task NOW:
Task(
subagent_type="team-shinchan:shinnosuke",
model="opus",
prompt=`/team-shinchan:autopilot has been invoked.
## Autonomous Execution Mode
Complete autonomously without user intervention:
1. Create WORKFLOW_STATE.yaml (stage: requirements)
2. Auto-analyze requirements with Misae
3. Create plan with Nene in MICRO-TASK FORMAT (REQUESTS.md, PROGRESS.md)
- Each phase broken into 2-3 minute tasks with exact file paths, code, and verification commands
4. Execute using micro-execute pattern (RULE 2.7 in agents/shinnosuke.md):
- For each micro-task: fresh implementer subagent → spec compliance review → code quality review
- See skills/micro-execute/SKILL.md for full protocol
5. After all execution phases complete, run Stage 4 (Completion):
- Write RETROSPECTIVE.md (summary, decisions, learnings) via Bo
- Write IMPLEMENTATION.md (overview, architecture, files changed) via Bo
- Action Kamen final verification of entire workflow
6. Auto-fix issues when discovered
## Stage Rule Compliance
- requirements Stage: Only requirements gathering (no code modification)
- planning Stage: Only planning (no code modification)
- execution Stage: Implementation proceeds
- completion Stage: Documentation and verification
## Safety Limits
- **Max iterations**: 15 (pause and report if reached)
- **Progress check**: If no measurable progress in 3 consecutive iterations, pause and report to user
- **Scope escalation**: If requirements analysis reveals 20+ files or 3+ domains, recommend switching to /team-shinchan:bigproject
## On Max Iterations Reached
If the 15-iteration limit is reached, output:
- Current stage and phase
- What was completed
- What remains incomplete
- Specific blocker (if any)
- Recommended next step for user
User request: ${args || '(Request content analysis needed)'}
`
)
STOP HERE. The above Task handles everything.
Source
git clone https://github.com/seokan-jeong/team-shinchan/blob/main/skills/autopilot/SKILL.mdView on GitHub Overview
This skill runs a complete workflow from requirements analysis through verification without user input. It orchestrates planning, execution, and documentation to deliver ready-to-verify results. It’s designed for auto, automatically, and autopilot requests with safety and scope controls.
How This Skill Works
On invocation, it validates input length and truncates if needed. It then orchestrates autonomous phases: requirements gathering, auto-analysis, and planning via micro-task formatting, followed by execution using a micro-execute pattern with subagents. After completion, it generates retrospective and implementation documentation and performs final verification, all under defined stage rules and safety limits for iteration and progress.
When to Use It
- When you want a fully autonomous workflow from requirements gathering to verification without human intervention.
- When handling repetitive, well-scoped requests that benefit from structured micro-task execution.
- When you need end-to-end delivery with generated documentation (RETROSPECTIVE.md, IMPLEMENTATION.md) and verification.
- When minimizing user input is a priority and you can rely on predefined process stages and safety limits.
- When a request fits within scope limits (e.g., avoids too many files or domains) and can be executed without mid-run changes.
Quick Start
- Step 1: Invoke the skill with your request via team-shinchan:autopilot (args).
- Step 2: The system validates input, then starts autonomous execution across requirements, planning, and execution phases.
- Step 3: Review the produced RETROSPECTIVE.md and IMPLEMENTATION.md and confirm verification results.
Best Practices
- Keep request scope narrow: avoid 20+ files or 3+ domains to prevent escalation.
- Limit input length to 2000 characters to ensure reliable autopilot processing.
- Define clear success criteria so the automation can generate verifiable outputs.
- Review generated RETROSPECTIVE.md and IMPLEMENTATION.md after completion for learnings and traceability.
- Monitor progress and be ready to pause or escalate if no measurable progress within safety limits.
Example Use Cases
- Auto-create a small feature from a user requirement and deliver a complete implementation plan with documentation.
- Autonomously analyze a bug report, plan a fix in micro-tasks, implement, test, and produce a retrospective.
- End-to-end onboarding of a new service: requirements → plan → code changes → verification → documentation.
- Auto-generate a deployment blueprint for a simple module and produce implementation notes.
- Auto-run a multi-step data-processing feature with required validations and post-run summaries.
Frequently Asked Questions
Add this skill to your agents
