reading-withholding

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell commands shinkoku pdf extract-text and shinkoku pdf to-image to process PDF documents. These tools are used to extract text or convert pages to images for subsequent analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted financial documents which creates a surface for indirect prompt injection. 1. Ingestion points: PDF and image files read via Read and shinkoku tools. 2. Boundary markers: Absent; the skill relies on structural output tags but does not explicitly instruct the model to ignore text within the documents. 3. Capability inventory: Access to shell commands (shinkoku) and file system tools (Read, Glob). 4. Sanitization: None; document content is processed as-is to extract fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 12:59 AM