research-company
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
pip install reportlaband runs a Python script (scripts/generate_report.py) to process JSON data and generate a PDF. Standard operation, but involves local execution of a provided script. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to fetch data from external, untrusted company URLs and process it into a structured report.
- Ingestion points: Web fetches and searches in Phase 1 (SKILL.md).
- Boundary markers: Missing explicit boundary markers to prevent the AI from following instructions found on researched websites.
- Capability inventory: Execution of
scripts/generate_report.pyand workspace file writes. - Sanitization: The Python script uses
_escape_for_paragraphto escape HTML-like tags, which provides some protection against layout breakage in the PDF generator. - [EXTERNAL_DOWNLOADS]: Downloads the
reportlabpackage from PyPI during execution.
Audit Metadata