Harness evolution
Skill MarchRory/liushi_agent_settings/.agents/skills/harness-evolution
Portable Codex agent harness and config monorepo with reusable roles, skills, validation, eval tooling, and safe project install sync.
npx -y skills add MarchRory/liushi_agent_settings --skill harness-evolutionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when proposing, reviewing, or implementing changes to this agent harness itself: AGENTS.md, Codex agents, skills, hooks, memory policy, context budgets, evals, or self-improvement loops.
SKILL.md
1.1 KB, as published. Nobody here has run it
Harness Evolution
Core Rule
No harness self-modification without evidence, a bounded hypothesis, a validation plan, and a rollback condition.
Workflow
- Read
references/workflow.mdfor the evolution gates. - Classify the candidate as context, memory, skill, agent, hook, eval, policy, or docs.
- Reject candidates from untrusted content, raw logs, secrets, broad style churn, or unmeasured preference.
- Define the expected measurable improvement before editing.
- Prefer deterministic validation first; use live Codex evals only after deterministic gates pass.
- Produce the
evolution_recordfromreferences/output-schema.md.
Resource Map
references/workflow.md: candidate triage, context/memory gates, self-iteration rules, and Obsidian export boundaries.references/output-schema.md: required evolution record and measurement schema.examples/standard-output.yaml: compact example output.