pdfcraft-cli-pages
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]:
scripts/bootstrap.pydownloads theprintcraft-clibinary from a GitHub release URL (github.com/storytold/pdfcraft). The process is secured by strict URL pattern matching and SHA256 checksum verification of both the downloaded archive and the final executable. - [REMOTE_CODE_EXECUTION]: The skill downloads and executes a remote binary to provide its core PDF editing functionality. The execution is mitigated by cryptographic integrity checks and source validation defined in
scripts/runtime.lock.json. - [COMMAND_EXECUTION]: The skill relies on executing system commands via
subprocess.runandsubprocess.Popenacross several files (scripts/bootstrap.py,scripts/ocr.py,scripts/command_gateway.py,scripts/verification.py,scripts/execution.py) to manage the CLI lifecycle and process document tasks. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests untrusted data from external PDF files during text extraction and OCR processes in
scripts/ocr.pyandscripts/verification.py. - Boundary markers:
SKILL.mdprovides instructions for the agent to perform manual verification of page counts, geometry, and text summaries to ensure output alignment. - Capability inventory: The skill has the capability to write to the file system (outputting modified PDFs) and execute native CLI commands.
- Sanitization:
scripts/execution.pyimplements a redaction mechanism that scans task parameters and outputs for sensitive keywords (e.g., password, token, api_key) and replaces them with[REDACTED]before returning data to the agent context.
Audit Metadata