skills/fastcrw/crw/crw-parse/Gen Agent Trust Hub

crw-parse

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions utilize shell commands to perform document parsing operations. It uses the crw CLI tool for local file scraping and the curl utility to transmit document data to the vendor's REST API at $CRW_API_URL.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external PDF documents, which represents a vulnerability surface for indirect prompt injection attacks.\n
  • Ingestion points: Untrusted data enters the agent context through local file paths (e.g., report.pdf) and base64-encoded PDF content provided via the contentBase64 parameter in the MCP tool.\n
  • Boundary markers: The skill does not define specific delimiters or "ignore instructions" markers to separate extracted document content from the agent's primary instructions.\n
  • Capability inventory: The skill environment permits file system access (Read) and network operations (curl), which could be leveraged if an agent is successfully manipulated by instructions found within a parsed document.\n
  • Sanitization: While the skill converts binary PDF data into structured text formats, it does not mention any sanitization or validation of the extracted text layer to filter out potential prompt injection payloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 09:37 AM