Skills
Navigate
skills/lyndonkl/claude/pdf-statement-parser/Gen Agent Trust Hub

pdf-statement-parser

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of PDF statements provided via the pdf_path parameter. An attacker could embed malicious instructions within a PDF (e.g., in a transaction description or hidden text) to influence the agent's behavior.
  • Ingestion points: The skill invokes the Read tool to ingest the content of local PDF files (SKILL.md).
  • Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between the document's data and potential instructions embedded within the PDF content.
  • Capability inventory: The skill uses the Read tool to access filesystem content. While it emits JSON, the description_raw field is preserved exactly as printed, which could propagate injection strings to downstream systems.
  • Sanitization: The skill normalizes dates and currency to integers, providing some validation, but lacks specific sanitization for text fields like description_raw or owner names.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:34 PM
Security Audit — agent-trust-hub — pdf-statement-parser