ClawSpec is now public.
It extracts the contract-first QA stack from openclaw-ceremonia into a standalone package for OpenClaw skills and agents:
clawspec validateforSKILL.mdandSOUL.mdclawspec runfor live scenario contractsclawspec initfor scenario scaffoldingclawspec coveragefor project-wide QA gaps
It ships with 19 assertion types, a Python API, an OpenClaw skill wrapper, and redacted real-world examples.
Repo: https://github.com/austinmao/clawspec
- Open-sourced
clawspec: contract-first QA for OpenClaw skills and agents. https://github.com/austinmao/clawspec - It gives one testing surface for structure (
SKILL.md/SOUL.md), scenarios, handoffs, and full pipelines. - Commands in v0.1:
validate,run,init,coverage. - It includes 19 assertion types, a Python API, and an OpenClaw skill wrapper.
- The repo also includes a redacted Ceremonia example set so the framework is grounded in real production contracts.
- If you are building multi-agent systems, the goal is simple: move QA from ad hoc smoke tests to explicit contracts.
Title: Open-sourced ClawSpec, a contract-first QA framework for OpenClaw agents and skills
Body:
I just published ClawSpec, a standalone QA framework extracted from production OpenClaw testing workflows.
The idea is to treat agent/skill testing as explicit contracts instead of scattered ad hoc scripts:
- structural validation for
SKILL.mdandSOUL.md - scenario contracts for single-target runs
- handoff contracts between stages
- pipeline contracts for orchestrators
- coverage reporting across a project
The initial release includes:
clawspec validateclawspec runclawspec initclawspec coverage- 19 shipped assertion types
- Python API
- OpenClaw skill wrapper
- example projects, including a redacted production-style showcase