csv-over-json-tables
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing untrusted data via the source_data block in the extraction template. This creates a surface for indirect prompt injection where malicious instructions in the source text could attempt to influence the CSV output. 1. Ingestion points: source_data block in SKILL.md. 2. Boundary markers: XML tags are used but instructions to ignore embedded content are missing. 3. Capability inventory: Provided Python and TypeScript logic for local parsing; no file, network, or command execution capabilities. 4. Sanitization: The TypeScript snippet lacks checks for reserved object properties, allowing a prototype pollution surface.
Audit Metadata