pdfcraft-cli-inspect
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the 'printcraft-cli' binary from the vendor's GitHub repository and an upstream repository (storytold/pdfcraft). These downloads are restricted to specific, verified URLs and are validated using hardcoded SHA256 hashes in 'runtime.lock.json' before any execution occurs.
- [COMMAND_EXECUTION]: The skill invokes its internal Python scripts and the downloaded CLI tool using the 'subprocess' module. It uses list-based arguments to avoid shell injection vulnerabilities while performing tasks like PDF rendering, text extraction, and OCR.
- [DATA_EXFILTRATION]: While the skill performs network operations to download its runtime, it includes a proactive 'secrets' detection and 'redact' mechanism in 'execution.py'. This system identifies and masks common sensitive keywords (e.g., 'password', 'api_key', 'token', 'credential') in the logs and receipts generated by the skill.
- [DYNAMIC_EXECUTION]: The skill uses 'importlib' to dynamically load its own helper modules (like 'execution.py' and 'bootstrap.py') from its distribution directory. This is used for internal modularity rather than executing external or untrusted code.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted PDF files. While it includes extensive verification steps (SHA256 checks, page count validation, geometry checks) to ensure the integrity of the processed output, it remains an ingestion point for external document content.
Audit Metadata