glmocr-table
Pass
Audited by Gen Agent Trust Hub on Mar 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious code, obfuscation, or deceptive patterns were found in the analyzed files.
- [DATA_EXFILTRATION]: The script scripts/glm_ocr_cli.py handles the ZHIPU_API_KEY via environment variables and communicates exclusively with the official ZhiPu AI endpoint at open.bigmodel.cn. This design prevents sensitive data from being sent to untrusted third-party servers.
- [COMMAND_EXECUTION]: All file and network operations are performed using standard, verified Python libraries such as requests. The script does not use any unsafe execution functions like eval, exec, or subprocess for untrusted input.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes external files and URLs. 1. Ingestion points: Local files or remote URLs via --file and --file-url. 2. Boundary markers: None explicitly defined in the output instructions. 3. Capability inventory: Local file read (path.read_bytes) and network POST requests to ZhiPu API. 4. Sanitization: None performed on the extracted OCR text before display. However, the risk is negligible given the agent is instructed to treat the output as a static Markdown table.
Audit Metadata