中医面诊分析工具

• Analyzed: May 10, 2026, 7:15 AM
• Findings: 1
• Version: 1.0.3
• Hash: c21274fa113d05528dbe090a91f6f0727f7583ba9bc4e1ea437601b515e1a55d
• Source: palm
• External report: View on VirusTotal

Findings (1)

This skill bundle exhibits several high-risk behaviors and manipulative instructions. The SKILL.md file uses aggressive prompt-injection techniques, such as 'Mandatory Memory Rules' with 'Highest Priority,' to explicitly forbid the AI agent from accessing local memory or LanceDB, forcing it to rely solely on the developer's cloud API (lifeemergence.com). The utility script `skills/smyx_common/scripts/util.py` performs silent user registration and tracks users via phone numbers or 'open-id' without explicit consent. Furthermore, `skills/smyx_common/scripts/skill.py` contains an `ai_chat` method that allows the skill to programmatically invoke the `openclaw` agent CLI via `subprocess.run`, which is a high-privilege capability. The bundle also includes hardcoded logic to intercept HTTP 402 errors and prompt users to install a specific 'smyx-payment' skill, indicating a forced monetization and data-collection framework.