菜单
Light Home Assistant Homelab work loop: find-work, pick with the operator, implement-change, review-loop, ship-work handoff, then STOP. Use when the operator says self-improve or wants one guided contribution lap.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Apocrathia/home-assistant-config --skill self-improve命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Mine Home Assistant history for automation trigger candidates from a narrative or structured event brief. Use when finding a reliable trigger, correlating a morning/evening sequence, ranking sensor signals for a new automation, or running the pattern-analyst ↔ pattern-reviewer pair.
Interview the operator about a topic until shared understanding, walking each branch of the decision tree one at a time. Use when stress-testing an idea, design, or requirement; resolving open questions before acting; or when the user says alignment, /alignment, or wants to get aligned.
Explore packages/, layout, and .agents/ for architectural friction (coupling, shallow packages, missing boundaries) and file or recommend agent issues under docs/issues/. Use when improving config shape, before large refactors, or when find-work should surface design-debt candidates.
基于 SOC 职业分类
正在显示 SKILL.md
| name | self-improve |
| description | Light Home Assistant Homelab work loop: find-work, pick with the operator, implement-change, review-loop, ship-work handoff, then STOP. Use when the operator says self-improve or wants one guided contribution lap. |
| disable-model-invocation | true |
One light lap: discover → pick (with operator) → implement → review → handoff → STOP. All edits happen in the main checkout.
No autonomous PR merging, no overnight ship loops, no merge cleanup via
clock-out, no watch-pr. Git stays operator-owned:
.agents/rules/operator-owned-git.md.
Work sources: .agents/context/work-sources.md
(incl. read-only .shopping_list.json).
flowchart TD
start[Session /self-improve]
start --> findWork["find-work (read-only)"]
findWork --> pick[Operator picks Launch N]
pick --> impl[implement-change]
impl --> review[review-loop]
review --> ship["ship-work (handoff only)"]
ship --> stop[STOP]
- [ ] 1. Run [`find-work`](../find-work/SKILL.md) (read-only)
- [ ] 2. Operator picks a launch brief (or names a task)
- [ ] 3. [`implement-change`](../implement-change/SKILL.md)
- [ ] 4. [`review-loop`](../review-loop/SKILL.md)
- [ ] 5. [`ship-work`](../ship-work/SKILL.md) handoff
- [ ] 6. STOP — do not loop, merge, or clock-out-merge-cleanup
Fuzzy briefs → alignment, then stop for proceed.
Protected paths → confirm before editing.
| Situation | Action |
|---|---|
| Operator names Launch N / a task | Do that |
| Operator only ran find-work | Stop after briefs |
| Vague shopping-list title | Alignment / scope first |
| Dedupe blocked or protected-path | Stop; surface why |
Do not walk briefs autonomously into commit/PR. Do not start the next find-work lap until the operator asks.
.shopping_list.json or .storage/ship-work as commit/push — it is handoff onlyclock-out as merge teardown (session wrap only, if asked)