soul-fireseed
PassAudited by VirusTotal on May 9, 2026.
Overview
Type: OpenClaw Skill Name: soul-fireseed Version: 2.1.0 The soul-fireseed skill bundle is a comprehensive personality modeling and memory analysis tool designed to extract cognitive and behavioral traits from user conversations. It utilizes keyword matching and semantic embeddings (via the sentence-transformers library) to build a structured 'persona model' and track its evolution over time. The codebase is professionally organized, featuring modular logic in lib/extractor.py and lib/distiller.py, and includes a full suite of unit tests. All file operations are restricted to a local user-data directory for storage and backups. While the SKILL.md instructions describe a 'background mode' for automatic extraction, this is presented as an opt-in feature requiring user consent. No evidence of data exfiltration, malicious execution, or harmful prompt injection was identified.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Private chat history could be converted into a reusable personality profile, including sensitive inferences about habits, emotions, relationships, and identity.
This directs the agent to review historical conversations and memories across sensitive life, emotion, relationship, health, and self-reflection topics without clearly bounding the scan scope.
AI 扫描已有记忆/对话记录 ... 检索用户历史对话中涉及以下内容的部分
Only use memory-analysis mode with an explicit time range and source list; require a preview before extraction, allow exclusions, and provide clear delete/export controls.
The agent may keep analyzing future conversations and updating the profile even when the user is not actively asking for analysis.
The skill explicitly describes ongoing background extraction that happens without the user's immediate awareness, which is persistent behavior beyond a single user request.
【每天模式】 → 每日对话末尾自动扫描 ... 在用户无感知的情况下完成提取
Default to manual mode, make every background extraction visible, provide an easy off switch, and log what was scanned and stored.
Sensitive inferred personality data may remain on disk and in backups after the conversation ends.
The default configuration persists extracted traits and personality profiles and creates backups, but the artifacts do not clearly document user-facing retention, deletion, or scope controls.
"fossil_storage_path": "user-data/fossils/", "persona_storage_path": "user-data/persona/", "auto_backup": true, "backup_retention_days": 30
Document storage locations clearly, add retention and deletion controls, and let users disable backups and embedding caches for sensitive use.
Installing unpinned dependencies can introduce unexpected package versions or supply-chain exposure.
The documented setup relies on external Python packages without exact version pins; this is purpose-aligned for embeddings and clustering but should be installed from trusted sources.
pip install sentence-transformers scikit-learn numpy pandas jinja2
Use a virtual environment, pin dependency versions, and install from trusted package indexes.