ClawHub
Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Self Learning Skill

v3.0.1

Agent 自我学习与记忆更新技能。分析对话历史，提取关键信息，自动更新配置文件和学习记录，实现 Agent 持续自我成长。 融合自学习 (配置文件更新) + 自改进 (学习记录系统) 双引擎。 Use this skill when: - 需要整理和更新 Agent 记忆 (MEMORY.md, IDENTIT...

⭐ 1· 1.9k·32 current·33 all-time
byA'c'c'z'd'y@acczdy
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Benign
medium confidence
✓
Purpose & Capability
Name/description (self-learning, memory/config updates) match the code and SKILL.md: scripts read conversation history, parse and update MEMORY.md/IDENTITY.md/TOOLS.md etc., and maintain a .learnings/ log. Allowed tools (Read/Write/Bash/SessionsList/SessionsHistory/SessionsSend) are appropriate for the stated functionality.
ℹ
Instruction Scope
Instructions and scripts will read and write core workspace files, create/append structured logs in .learnings/, back up/restore files, and install a hook into ~/.openclaw/hooks that injects suggestions into sessions. This is within scope but grants the skill potential to modify agent behavior and to read an OpenClaw config at /root/.openclaw/openclaw.json for workspace auto-detection — users should be aware that the skill will access those files and session history.
✓
Install Mechanism
No external install/downloads or remote code fetches are present. There is no install spec; the package is instruction + local Python scripts. requirements.txt only lists standard Python packages (PyYAML, pytest, black, flake8).
ℹ
Credentials
The skill declares no required environment variables or credentials, and that matches registry metadata. At runtime it does read WORKSPACE (if present) and may read OpenClaw config files (e.g., /root/.openclaw/openclaw.json) for auto-detection. No network exfiltration endpoints or secret-collection code were found, but reading an agent config file could expose workspace paths or settings — consider whether that access is acceptable.
ℹ
Persistence & Privilege
always:false (no forced inclusion). The skill installs a local hook into the user's OpenClaw hooks directory and the hook can inject messages into sessions (onSessionStart/onPromptSubmit). This is coherent for a learning skill but elevates its influence over agent interactions — review hook behavior before enabling. The skill does not modify other skills' configs.
Assessment
This skill is coherent with its stated purpose (automatically reading conversation history and updating memory/config files and logs), but it does perform filesystem writes and can install a session hook that injects messages. Before installing: 1) Review and back up your workspace (MEMORY.md, IDENTITY.md, etc.). 2) Inspect scripts/memory_update.py and hooks/openclaw/handler.js to confirm the exact files/paths it will read and write (it attempts to read WORKSPACE and /root/.openclaw/openclaw.json for auto-detection). 3) Note the hook behavior — enable only if you trust the injection messages. 4) Check config.yaml (safety.require_confirm_for_delete defaults differ from some doc text) and set require_confirm_for_delete:true if you want stricter deletion safeguards. 5) Run in dry-run/preview mode first (python3 scripts/memory_update.py --dry-run) to observe proposed changes before allowing writes.

Like a lobster shell, security has layers — review code before you run it.

latestvk97e7nctbea1ye319tvmxwv0x182bs4q

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments