Help us improve
Share bugs, ideas, or general feedback.
From ai-inner-os
Distills user profiles from prompt history or local agent chat logs, extracting personality, work traits, and communication preferences. Opt-in for explicit requests with privacy safeguards.
npx claudepluginhub summersec/ai-inner-os --plugin ai-inner-osHow this skill is triggered — by the user, by Claude, or both
Slash command
/ai-inner-os:user-profile-distillationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill is opt-in only. Do not use it unless the user explicitly asks for user profiling, personality/work-style analysis, prompt-history distillation, continuous profile evolution, or similar wording.
Analyzes Claude session data for behavioral dimensions, extracts strengths and their structural costs (shadows), and generates attitude principles with practice matrix reports.
Builds and manages a cognitive profile from session history — work patterns, blind spots, cross-domain connections. Use for self-reflection or session startup.
Interviews user on stable cross-project preferences like communication style, tone, technical depth, and environment defaults, saving conclusions to Honcho memory.
Share bugs, ideas, or general feedback.
This skill is opt-in only. Do not use it unless the user explicitly asks for user profiling, personality/work-style analysis, prompt-history distillation, continuous profile evolution, or similar wording.
Do not read local history, transcripts, databases, or cached prompts unless the user explicitly asks you to do so for this profile task.
Continuous evolution is also opt-in. Do not silently update a profile across turns unless the user explicitly asks to keep refining, evolve, maintain, or update the profile during the conversation.
Use one of these sources:
agent-chat-history skill to extract user prompts.For local history, prefer a bounded date range. If no date range is provided, ask for one before reading history.
Recommended command from the bundled agent-chat-history/ skill directory:
python scripts/query_history.py --date YYYY-MM-DD --prompts-only --json
Use --mode claude, --mode codex, or --mode cursor when the user limits the source.
Use this mode only when the user explicitly asks for the profile to keep evolving during the current conversation.
The goal is not to create permanent memory. The goal is to maintain a visible, revisable working profile for the current task or conversation, similar to a self-improving feedback loop:
Profile v1 from approved inputs.confirmed, likely, or tentative; ask the user to correct anything that feels wrong.Profile v2, with a short change log.When continuous evolution is enabled, append this section after the normal profile:
## 持续进化记录
- 当前版本:
- 本轮新增证据:
- 画像增量:
- 被修正/降级的判断:
- 下一轮协作策略:
- 需要用户确认:
Use this structure:
## 数据范围
- 来源:
- 时间范围:
- 样本量:
- 置信度:
## 工作画像
- 常见任务:
- 关注重点:
- 质量标准:
- 决策方式:
## 沟通与协作偏好
- 指令风格:
- 反馈偏好:
- 对 AI 的期待:
- 容易不满意的点:
## 表达特征
- 语言习惯:
- 组织方式:
- 情绪/语气特征:
## 协作建议
- 回答应如何组织:
- 需要主动提醒什么:
- 应避免什么:
## 不确定性与边界
- 样本偏差:
- 不能推断的内容:
- 需要用户确认的问题:
## 持续进化记录(仅在用户开启持续进化时输出)
- 当前版本:
- 本轮新增证据:
- 画像增量:
- 被修正/降级的判断:
- 下一轮协作策略:
- 需要用户确认:
Be direct, respectful, and non-clinical. Describe working preferences and interaction patterns, not fixed identity. Prefer "seems to", "often", and "based on this sample" over absolute claims.