AI 树德:义商本体伦理安全系统

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

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.

NoteHigh Confidence
ASI01: Agent Goal Hijack
What this means

The agent may decline, reframe, or modify answers according to this skill's value-alignment rules.

Why it was flagged

The skill explicitly instructs the agent/system to refuse or recalibrate outputs under its ethics framework. This is central to the stated safety purpose, but it can alter normal response behavior.

Skill content
当检测到以下情况时,系统将自动触发防护... 🚫 **拒绝机制**... 🌈 **价值观校准**
Recommendation

Install only if you want this ethics framework to influence agent behavior; review the stated triggers and refusal criteria.

What this means

If the user runs the setup command, they rely on whatever current package versions are retrieved from the package index.

Why it was flagged

The setup instructions ask the user to install unpinned Python packages. This is user-directed and not auto-executed, but dependency versions and provenance are not constrained.

Skill content
pip install regex numpy pandas
Recommendation

Use a trusted environment and consider pinning or reviewing dependency versions before installing.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

If started, the monitor may keep running and consuming local resources until the user stops it.

Why it was flagged

The monitor can start a background thread and run a loop until stopped. This is consistent with a monitoring feature and is only shown as user-started.

Skill content
self.thread = threading.Thread(target=self._monitor_loop, args=(interval,), daemon=True) ... while self.running:
Recommendation

Start the monitor only intentionally, avoid configuring it for automatic startup unless needed, and stop it when finished.

What this means

Sensitive text included in monitored responses could appear in local logs.

Why it was flagged

The monitor logs response snippets and can write logs to disk. This is local and purpose-aligned, but monitored text excerpts could persist in log files.

Skill content
self._log(f"检测响应:{response[:50]}...", "INFO") ... with open(path, 'w', encoding='utf-8') as f:
Recommendation

Choose log locations deliberately, avoid monitoring sensitive content unless necessary, and delete logs when they are no longer needed.