ai-to-human-zh
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a local Python script (
python3 scripts/hanyihan_qc.py) to perform text analysis. The script performs harmless regex-based pattern matching on local text files to identify common AI-generated text patterns.- [REMOTE_CODE_EXECUTION]: The README mentions an installation command (npx -y skills add Holden323/ai-to-human-zh -g --all). This is the standard installation method for the platform and is considered a safe operational procedure for skill deployment.- [PROMPT_INJECTION]: The skill uses phrases like 'Iron Rule' and 'Must obey,' but these are framed within the context of the skill's specific workflow (e.g., generating a report before making changes). These do not attempt to bypass core AI safety filters or override global system instructions.- [DATA_EXFILTRATION]: The Python script and instructions are limited to local text analysis. No network operations, sensitive file access, or credential harvesting patterns were observed.- [INDIRECT_PROMPT_INJECTION]: While the skill processes external text data, it includes strong boundary markers and a 'human-in-the-loop' workflow requirement, where the agent must present a report and obtain user confirmation before performing any modifications. This significantly reduces the risk of the agent autonomously following malicious instructions embedded in the analyzed text.
Audit Metadata