knowledge-base-working-profile
Skill8 starsChanged 7 months agoMITView on GitHub
This skill should be used when the user asks to remember or update how they work, distill collaboration preferences from ongoing conversations, maintain a working profile in a markdown/wiki knowledge base, 沉淀协作画像, 更新我的工作偏好, 记录我的决策习惯, or 从沟通过程中整理长期画像. It extracts stable collaboration-relevant signals such as preferences, decision heuristics, boundaries, and anti-patterns, keeps confirmed and inferred items separate, and refuses to store secrets or overly sensitive personal data.
In this repository9 other files it gives its agents.
The file
Files
Knowledge Base Working Profile
把持续沟通中出现的稳定协作信息沉淀进 markdown/wiki 知识库。
它处理的不是“原始个人资料”,而是未来协作真正会反复用到的东西,例如:
- 稳定事实(仅限工作相关)
- 偏好
- 决策习惯
- 协作边界
- 反模式
- 仍待确认的画像信号
它和 knowledge-base-maintenance 的区别是:
knowledge-base-maintenance:沉淀某次任务/聊天/排障的结果knowledge-base-working-profile:沉淀这个人长期怎么合作、怎么判断、偏好什么、不喜欢什么
默认目标不是“做人格克隆”,也不是“存个人隐私档案”,而是形成一个可持续更新的 working profile,减少反复解释与协作摩擦。
Required input
执行前先读取:
references/working-profile-schema.mdreferences/evidence-thresholds.mdreferences/sensitivity-boundaries.md../../templates/vault-profile-template.md../../templates/working-profile-page-template.md
默认需要这些输入:
- vault profile
- 当前 working profile page(如果已有)
- 本轮证据来源:当前对话、用户纠正、长期反馈、会议纪要、任务回顾等
- 如果 profile 里没有 working profile page:用户指定目标页
如果没有明确 working profile page,不要猜测文件路径。
Core rules
- 只沉淀对未来协作有用的信息。
- 先过敏感信息过滤,再判断是否入库。
- 明确区分 confirmed / repeated / inferred。
- 单次随口表达不等于稳定画像。
- 允许时间变化和场景差异,不强行写成绝对人格标签。
- 优先更新已有 profile,不频繁新建画像页。
- 不写 secrets,不写高敏感个人信息,不写第三方隐私。
- 重大画像变化应显式写入 change note,而不是悄悄覆盖。
What belongs in a working profile
Good candidates
- 经多次重复出现的工作偏好
- 用户明确表达过的协作方式偏好
- 决策习惯与推进节奏
- 对输出风格、结构、颗粒度的稳定要求
- 明确的边界和禁区
- 常见不满意点或反模式
Bad candidates
- 一次性情绪
- 临时状态
- 无法验证的私人推测
- 无助于未来合作的个人细节
- 账号、密码、证件、地址、财务账户、医疗隐私等高敏感信息
Recommended workflow
1. Gather evidence first
先收集本轮真正可作为证据的材料:
- 当前对话中的明确表达
- 用户对你方案的修正
- 用户多次重复过的偏好
- 过往 working profile 中已有条目
- 近期任务里出现的新信号
2. Run the sensitivity gate
先按 references/sensitivity-boundaries.md 过滤:
- 哪些内容绝不应入库
- 哪些内容需要降级成模糊描述
- 哪些内容只适合短期上下文,不适合长期页面
3. Extract candidate signals
把证据转成候选画像信号,而不是直接照抄原话。
优先抽这些类型:
- Stable facts
- Preferences
- Decision heuristics
- Collaboration boundaries
- Anti-patterns
- Open questions / provisional signals
4. Assign evidence strength
使用 references/evidence-thresholds.md 的口径:
confirmed:用户明确确认repeated:多次稳定出现inferred:当前可推断,但还不够稳
inferred 项默认应更克制,必要时放到 provisional 区域,而不是写死。
5. Update the working profile page
默认按 ../../templates/working-profile-page-template.md 的结构更新。
写入时:
- 保留对未来协作真正有用的表述
- 压缩原始上下文
- 避免人格化夸张描述
- 保留必要 change notes
- 避免同义重复
6. Sync visibility surfaces carefully
根据 profile 中的 working profile visibility 决定同步范围:
maintainer-only:只更新 profile 页面和维护入口reader-visible:可同步 reader entrypoint 或治理入口- 如果未定义 visibility,默认按
maintainer-only处理
7. Validate before finishing
检查:
- 有没有误写高敏感信息
- 有没有把一次性情绪写成长期标签
- 有没有把 inferred 写成 confirmed
- 是否保留了变化轨迹
- 页面是否仍然可维护、可扩展、可更新
Output expectations
汇报时至少说明:
- 新增或更新了哪些 working profile 条目
- 哪些属于 confirmed / repeated / inferred
- 哪些候选信息被过滤掉,为什么
- 是否记录了 change note
- 是否同步了对应入口页
Resources
references/
references/working-profile-schema.md:working profile 推荐结构与字段边界references/evidence-thresholds.md:如何判断 confirmed / repeated / inferredreferences/sensitivity-boundaries.md:哪些个人信息绝不应入库,哪些应降级处理
templates/
../../templates/vault-profile-template.md:working profile page 与 visibility 等配置入口../../templates/working-profile-page-template.md:推荐的 working profile 页面结构
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

