JasonxzWen/harness-hub

README preview

# Harness Hub

[简体中文](README.zh-CN.md) | English

Harness Hub is a personal repo harness toolkit for making agent work repeatable across projects. It installs the target-distributed standard skill/routing set, initializes the standard target harness when requested, validates the result, and keeps managed files safe through lock-backed lifecycle commands.

Imported skills can keep their upstream style; Harness Hub mainly owns routing, source records, harness templates, and lifecycle safety.

Agent execution rules live in canonical [AGENTS.md](AGENTS.md); [CLAUDE.md](CLAUDE.md) imports it for Claude Code. Human-facing workflow detail lives in [Development Workflow](docs/development-workflow.md), with delegated-agent acceptance and arbitration patterns in [Agentic Loop Catalog](docs/agentic-loop-catalog.md).

## First-Use Summary

Harness Hub does a small set of bounded jobs:

- `check` and `analyze` inspect a target repo and stay read-only.
- `init-harness --target standard` creates the root harness only when you explicitly approve it.
- `install` installs the target-distributed standard skill set only; it does not create root harness files.
- `standard` migrates prompt, context, harness, and loop engineering resources plus reusable skills, while Harness Hub source-maintenance resources stay local to this repository.
- the standard harness includes an LLM Wiki context pack under `.harness-hub/context/`.
- `loop evaluate` and `loop schedule` decide continue vs interrupt and can append local Loop ledgers after `--yes`; `loop required` and `loop verify` derive dirty-path or base/head review gates and block handoff when loop evidence is missing; `loop run-start`, `agent-record`, `lease-check`, `collect-trace`, and `integrate` record local subagent orchestration state under ignored `.harness-hub/state/runs/`.

View full repository on GitHub →