Autoharness — Development skill for Claude Code
macOS coding-agent control center: one objective, routed to direct execution, a bounded loop, a swarm, or a dynamic DAG.
How to install Autoharness
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open codejunkie99/autoharness and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Autoharness does
macOS coding-agent control center: one objective, routed to direct execution, a bounded loop, a swarm, or a dynamic DAG. Local only, no external writes.
Alternatives in Development
- Gyoshu Auto — Start goal-based autonomous research with bounded execution 244 ★
- Omk — Evidence-gated runner for Codex, Claude Code, OpenCode, and local coding agents 141 ★
- Ccc Orchestration — Spawn, inject into, and ask questions of persistent sibling sessions via Claude Command Center (CCC) 130 ★
README
AutoHarness
[](#requirements) [](#requirements) [](Cargo.toml) [](LICENSE) [](#test) [](SECURITY.md)
Badges are declared facts, not CI output. There is no hosted pipeline: the GPUI shell needs a Metal toolchain, so the gate in [CONTRIBUTING.md](CONTRIBUTING.md) runs locally and is what those numbers come from.
A macOS-only coding-agent control center. Accepts one objective through coordinator chat, routes it to direct execution, a bounded loop, a swarm, or a dynamic DAG, and orchestrates installed Codex or Claude CLI sessions — all local, no external writes. See `PLAN.md` for the full product plan, and [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) for how the pieces fit.
**Status: all local V1 phases are complete in code.** Foundation, engine bridge, sandbox, durable queues, direct chat, router and loops, graphs and swarms, replayable review surfaces, memory/evolution, the true-black cockpit, and release hardening ship together. Release hardening covers data integrity, export, privacy purge, diagnostics, semantic keyboard/VoiceOver accessibility, native attention, universal packaging, and a fail-closed signed-update path. Production signing, notarization, and update-feed hosting still require the owner's external credentials. A raw interactive PTY is intentionally outside AutoHarness's structured-agent scope; captured ANSI check output and classified unified diffs are built in.
Related Skills
Pursue
Hand Kimi an objective to pursue AUTONOMOUSLY across multiple turns (experimental goal mode), bounded by a har
Cold Eye Skill
Claude Code skill: objective, expert-level output judged against an external measurable bar (not the asker's t
Swarm Pi Code Plugin
connects Claude Code and Codex to a bounded Pi coding worker for repository-grounded analysis, planning, revie
Oh My Multica
Production-grade software delivery for coding-agent teams on Multica — dynamic planning, deterministic executi
Agent Control Center
Agent 控制中心:Flutter 手机端通过 Happier Relay(E2EE)远程控制电脑上的 Codex / Claude Code / OpenCode / OpenClaw / Hermes
App Intents Agent Skill
Agent skill that helps Claude Code, Codex, Gemini, and Cursor write correct Swift App Intents code for Siri, S
Related Agents
A01 Orchestrator
A01 ORCH — Swarm control plane. Use to decompose a brief into a task DAG, schedule/assign tasks to the other 1
Project Structure Reviewer
Read-only audit of the current repo against spec/project/project-structure/, producing a severity-sorted findi
X402 UI Designer
Visual Design & Layout Agent for x402 UI/Dashboard Swarm. Specializes in multi-layer parallax scroll systems,