hill-climb
Improve something step by step against a score until it stops getting better. Use when I say "hill climb".
- License 许可
- MIT
- 最近更新
- 2026-11-12
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- hill-climb
- description
- Improve something step by step against a score until it stops getting better. Use when I say "hill climb".
- disable-model-invocation
- true
02 资源文件 1
- 参考文档agents/openai.yaml43 B
03 SKILL.md 正文
Hill climb
Hill climbing uses agents to keep improving one or more metrics: measure, look for what could make it better, change that, measure again to see if it did, and loop. The score keeps the agents honest, so the climb is only as good as the metrics. Pick ones that move when the thing really gets better, and watch that nothing else gets worse along the way.
If my prompt leaves the metric or when to stop unclear, suggest them and wait for my go.
Metrics come in two kinds:
- Measured ones like load time, bundle size or a benchmark: run them the same way each time (
benchmark-checklist) and ignore gains inside the noise. - Judged ones like UI/UX feel or writing: a rubric scored by a council of models from my subscriptions (the ones I name, or the best
orchestrator_capabilitieslists), each scoring on its own. Score the best apps in the niche the same way, so you know what good looks like.
Keep a log of the rounds in the repo and show me how the score moved at the end.
04 同仓库更多 Skills
writing-for-agents
Write docs that agents read (skills, AGENTS.md, CLAUDE.md). Use when creating or editing a skill, or changing AGENTS.md or CLAUDE.md.
unslop
Make prose sound human, not AI-written. Use whenever you write a PR description, commit message, doc, README, release note, post, email or UI copy, or when asked to unslop or de-AI text.
tutor
Run a study session from my exam plan and progress files in the study repo. Use when I say study, teach me, quiz me, practice for an exam, or check whether I really understand a topic.
threads
Start, steer and read T3 threads on my laptop or Mac mini from either machine. Use to move a thread to the Mac, get a Windows check from the laptop, manage threads on both, or answer a message from the other machine.
statechart
Check a TLA+ model and draw it as a Mermaid statechart, with the shortest path to each broken rule. Use for /statechart, to read a TLC result, or to show a bug as a path through states.
sol-loop
Review a change with GPT Sol in a loop until we both agree. Use when I say "sol loop".