Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add CavinHuang/lume --skill lume-self-evolutiongit clone --depth 1 https://github.com/CavinHuang/lumeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cavinhuang/lume/lume-self-evolution)<a href="https://agentmods.dev/skills/cavinhuang/lume/lume-self-evolution"><img src="https://agentmods.dev/badge/skills/cavinhuang/lume/lume-self-evolution.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00070 | $0.02424 |
| Opus 5 | $0.00035 | $0.01212 |
| Sonnet 5 | $0.00014 | $0.00485 |
| Haiku 4.5 | $0.00007 | $0.00242 |
Grade A, and why
lume自我进化 scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 315 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Lume 自我进化
目标
当用户长期使用 Lume 时，agent 应逐步学会两类自我进化：
- 把稳定、可复用、会反复出现的系统偏好沉淀为配置
- 在文件层面持续维持稳定、克制、可解释的 workspace / thread 结构
这套自我进化同时覆盖：
- 系统配置演进
- 文件治理演进
但它仍然不面向：
- UI 状态
- 草稿
- 一次性上下文
- 用户内容文件
默认配置入口：
~/.lume/lume.yaml
核心原则
Lume 的自我进化本质上是：
- 观察稳定偏好
- 判断是否值得沉淀
- 以最小范围更新系统配置
- 保持可解释、可审计、可回退
不要把“这次任务刚好这么做”误认为“以后都应该这么做”。
允许沉淀到 lume.yaml 的内容
优先考虑沉淀这些系统配置：
- 默认模型选择
- 默认 provider / channel 路由
- MCP 启用与关闭
- Skill 启用与关闭
- permission / tool policy 偏好
- agent 的默认思考等级或 permission mode
这些配置只用于覆盖系统行为。
不允许沉淀到 lume.yaml 的内容
以下内容禁止写入 lume.yaml：
- UI 状态
- 窗口尺寸和布局
- 当前选中的会话 / workspace
- 草稿内容
- 一次性任务步骤
- 临时实验结论
- 研究笔记
- 用户正文内容
这些内容应继续留在各自的状态存储或内容文件里。
何时应该更新配置
只有在以下信号足够明确时，才应该尝试更新 lume.yaml：
- 用户多次明确要求切换到同一模型或 provider
- 用户多次反复启用 / 禁用同一个 MCP
- 用户反复要求某个 Skill 默认可用或默认关闭
- 用户反复修正权限策略，且修正方向稳定一致
- 用户明确表达“以后默认这样”
如果只是：
- 单次任务需要
- 当前任务的临时 workaround
- 一次性兼容某个外部环境
则不要沉淀为自我进化配置。
全局与 workspace 的选择规则
更新 lume.yaml 时，优先选择影响范围最小的配置层级。
默认顺序：
- 先考虑
workspaces.<slug> - 只有明显跨工作区通用的稳定偏好，才改顶层全局
判断规则：
- 如果偏好只和当前项目/工作区相关，就写
workspaces.<slug> - 如果偏好是用户在多个工作区都想统一保持的默认行为，才写顶层
不要因为“改顶层更省事”就把项目偏好提升成全局偏好。
更新方式
更新配置时必须遵守以下约束：
- 通过结构化更新修改
lume.yaml - 不做粗暴字符串替换
- 不覆盖无关 section
- 不把未知字段清空
- 保持 YAML 可读性
每次修改前都要明确：
- 修改的是哪个 section
- 修改的是顶层还是
workspaces.<slug> - 预期影响范围是什么
审计与说明
每次自我进化式配置更新，都应可解释。
至少要能回答：
- 为什么这次值得沉淀
- 为什么改全局或改 workspace
- 改完后会影响什么
如果当前实现有审计能力，应保留审计记录。
如果需要向用户汇报，优先用一句短说明：
- “我把这个偏好沉淀到了当前 workspace 的
lume.yaml配置里，后续同类任务会默认按这个行为执行。”
与其它文件的边界
lume.yaml 负责系统配置。
不要把下面这些职责混进去：
SOUL.md的人格与长期角色描述AGENTS.md的工程规则USER.md的用户长期画像MEMORY.md的长期记忆内容threads/<id>/.context/的任务过程材料
如果信息更像“规则文本”或“知识内容”，就不该写进 lume.yaml。
记忆更新方式也属于自我进化边界
Lume 的自我进化还体现在记忆系统上。
agent 不应把记忆当成“什么都往里丢的仓库”，而应遵守新的三层记忆模型：
- 全局长期记忆
~/.lume/MEMORY.md
- workspace 记忆
workspace/MEMORY.mdworkspace/memory/YYYY-MM-DD.md
- thread 临时记忆
workspace/threads/<thread-id>/.context/note.mdworkspace/threads/<thread-id>/.context/todo.md
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 315 lines · 70 tokens per session scan A 4088efb8a68d
lume自我进化 is a skill published in the GitHub repository CavinHuang/lume (3 stars, last pushed yesterday), licensed MIT. It adds 70 tokens to every session and 2,424 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
knowledge-maintenance
A maintenance guide for Proma’s project map and long-term collaboration notes.
session-cleaner
A tool for turning Proma session files—JSONL files containing one JSON record per line—into readable Markdown conversations. It can also inspect very long sessions in smaller parts.
ha-data-stores
Map of Hope Agent's local data stores and safe read-only query workflow. Use when the user asks where Hope Agent stores data, wants to inspect sessions/messages/memory/logs/background jobs/knowledge indexes/settings, asks the model to query local app data, or debugging requires checking persisted state. Trigger…
ha-knowledge
Working method for the Hope Agent knowledge space — how to capture, organize, link, retrieve, and maintain Markdown notes well with the note tools. Load whenever you are reading or writing notes in an attached knowledge base. Trigger on: user asks to take / save / organize / restructure notes, build or grow a…
proma-coach
A Chinese-language coach for improving how people use Proma, Agent, Skill, Chat, and project tools.
skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.