glmocr-table
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses environment variables (
ZHIPU_API_KEY) for secure credential management and communicates exclusively with the official ZhiPu AI endpoint (https://open.bigmodel.cn/api/paas/v4/layout_parsing), mitigating risks of credential exposure to unauthorized third parties. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external files (images and PDFs) and returns extracted Markdown text to the agent. This represents a potential indirect prompt injection surface if the extracted content contains malicious instructions designed to influence the agent's future behavior.
- Ingestion points: Untrusted image or PDF files and URLs provided as input to
scripts/glm_ocr_cli.py. - Boundary markers: None identified in the output formatting instructions to encapsulate the extracted text.
- Capability inventory: The CLI script has the ability to read local files and make network requests to the ZhiPu API.
- Sanitization: No evidence of input validation or output sanitization for the extracted text content.
Audit Metadata