split-pdf

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill utilizes curl, wget, and WebFetch to download PDF documents from external URLs discovered through web searches or scholarly search tools.
  • [COMMAND_EXECUTION]: The instructions involve executing shell commands for package management (uv pip install PyPDF2) and invoking external CLI utilities such as scholarly and paperpile to retrieve research metadata and documents.
  • [REMOTE_CODE_EXECUTION]: The skill requires the installation of the PyPDF2 package from a public registry to perform PDF manipulation tasks.
  • [PROMPT_INJECTION]: The skill processes untrusted PDF data downloaded from external sources. An attacker could embed instructions within a PDF to influence the agent's behavior during the reading and extraction process.
  • Ingestion points: External PDFs downloaded via search or direct links in Step 1.
  • Boundary markers: The skill uses a structured 8-dimension extraction framework and isolates reading in a subagent, but does not include explicit instructions to ignore embedded commands within the PDF text.
  • Capability inventory: The subagent has access to Python execution, file writing, and shell commands through its parent process context.
  • Sanitization: No content filtering or sanitization is performed on the extracted PDF text before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 06:17 AM
Security Audit — agent-trust-hub — split-pdf