Soul Archive
Paste the following prompt into your AI chat to install this skill:
Please install @user_dfcfbf47/soul-archive according to https://skillhub.cn/install/skillhub.md.
About this skill
Problem It Solves
AI agents often lose a user’s language habits, viewpoints, workflow preferences, and hard-won corrections across sessions. A single prompt can steer one answer, but it does not accumulate a durable profile of who the user is, how they like to work, or which mistakes were made before. Soul Archive targets this gap by turning confirmed conversational signals into local, maintainable memory that later agents can read, role-play, or use before acting. It treats persona and memory as local data assets rather than relying on cloud upload or one-off long prompts.
How It Works
Soul Archive separates skill behavior from stored data. Skills provide modes and extraction logic, while JSON records live under ~/.agent-guild/skills_data/soul-archive/ and are controlled by config.json. The main workflow is:
- Soul Extract: capture high-confidence signals via trigger phrases or auto_extract, covering identity, personality, language, knowledge and views, episodic memory, workflow preferences, and aspirations. Similar entries are merged; conflicting entries are flagged rather than silently overwritten.
- Soul Chat: build a role-play system prompt from the archive and let a digital clone converse in the user’s style; if directly asked whether it is AI, it must disclose that it is AI.
- Soul Report: generate an interactive HTML profile with radar charts, word clouds, topic heat maps, timeline views, and conflict views.
- Context Injection / Agent Memory / Self-Improvement: inject a compact persona summary into other agents, recall past corrections before tasks, and record corrections, reflections, and distilled patterns after tasks.
Boundaries And Caveats
It fits local, long-term, auditable persona memory, especially when multiple local IDEs or agents need the same profile. Use caution when auto_extract is enabled because the AI will extract persona signals during conversation; sensitive topics default to user confirmation. Data is not uploaded, but whether the Soul Chat prompt reaches an external LLM depends on the user’s agent or platform configuration. It should not be treated as a system that invents memories the user never expressed, nor as an automatic resolver for conflicting records.
Use Cases
- Want local IDE agents to load your workflow preferences, hard rules, and output style at session start instead of repeating context.
- Train an agent to avoid repeated mistakes by recording user corrections, command failures, and stale knowledge, then distill reusable patterns.
- Extract language style, viewpoints, and episodic memory from past conversations into reviewable persona profiles, topic heat maps, and conflict reports.
- Run role-play conversations with a digital clone while keeping the user’s tone and requiring honest AI disclosure when directly asked.
Best For
- Engineers maintaining multi-agent workflows who need to persist preferences, hard rules, and output styles in a shared local archive.
- Users long-term training AI assistants who want agents to remember failures and recall relevant corrections before acting.
- Privacy-sensitive users who need persona data stored locally and dimension-level extraction controlled through config.json.
- Developers building role-play or clone conversations who need system prompts grounded in real archive data and honest AI disclosure.
Related Skills
Local workflow memory with matching and SOP updates.
An OpenClaw live streaming executor that initializes TRTC streaming, starts a real-time dashboard, generates viewer URLs, and continuously reports live events.
Breaks down physical supply chains for super-trends to identify second- and third-layer bottlenecks, runs valuation and reverse checks, and maintains trackable reports.
A token-saving compression mode for Chinese LLMs with lite, full, ultra, and classical tiers, preserving code and technical terms while handling edge cases.