autopilot

Automate harness-infra issue selection and PR delivery via ship-spec.

28|3|Updated Mar 27, 2026
One-click install
npx skills add https://github.com/mifunedev/openharness --skill autopilot-mifunedev
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autopilot
Source: https://github.com/mifunedev/openharness/tree/main/.claude/skills/autopilot
Command: npx skills add https://github.com/mifunedev/openharness --skill autopilot-mifunedev

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Autopilot coordinates unattended harness-infra improvement cycles by selecting the oldest open autopilot issue or generating a first-principles research ticket, then driving end-to-end through the ship-spec pipeline to land a ready-for-review PR. It operates entirely on harness-infra tasks and runs in its own per-run tmux session, enforcing caps to prevent overload and providing clear session lifecycle management.

Core Features & Use Cases

  • Queue-driven issue selection: picks the oldest actionable autopilot-labeled issue with no linked PR.
  • First-principles research branch: if queue is empty, runs harness-audit to file a research ticket and builds a plan via /ship-spec.
  • End-to-end build pipeline: /ship-spec owns implement/eval/pr-audit and creates a ready-for-review PR, with a selection rationale and session lifecycle management.
  • Safe concurrency and governance: caps limit number of open/autopilot PRs per UTC day and total open; dry-run mode to preview decisions.
  • Session and artifacts management: tmux sessions named after the work branch, memory logs, and restore-safeguards.

Quick Start

Configure the autopilot cron to monitor an autopilot-labeled queue and let it autonomously file issues, build plans with /ship-spec, and create ready-for-review PRs.

Frequently Asked Questions about autopilot

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate end-to-end governance of GitHub infrastructure issues?

Automated infrastructure governance selects the oldest open labeled issue, builds a plan, and delivers a ready-for-review PR. It runs in an isolated tmux session, enforces concurrency caps, and documents a selection rationale.

Can I automatically generate research tickets when my infrastructure issue queue is empty?

When the issue queue is empty, a first-principles research branch is generated by running an audit to file a new research ticket. It then builds an implementation plan via the pipeline.

How do I limit concurrent open PRs for automated infrastructure workflows?

Concurrency is managed by enforcing caps on the number of open and autopilot PRs per UTC day and total open PRs. A dry-run mode previews decisions to prevent overload.

Does the ship-spec pipeline handle implementation, evaluation, and PR audits?

The ship-spec pipeline owns the implement, evaluate, and pr-audit stages. It creates a ready-for-review PR while satisfying planning and traceability requirements.

How do I run multiple autonomous infrastructure tasks without session conflicts?

Each run operates in its own per-run tmux session named after the work branch. This prevents conflicts and provides clear session lifecycle management and restore-safeguards.

What is the best way to autonomously drive infrastructure improvement cycles from issue to PR?

Autonomous improvement cycles are driven by monitoring a labeled issue queue via cron. It selects actionable issues with no linked PR and runs the pipeline to land a reviewable PR.