research-external

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local Python scripts (nia_docs.py, perplexity_search.py, firecrawl_scrape.py) for data retrieval. These scripts are invoked with parameters derived from user input, which are handled using standard shell quoting to reduce risks.
  • [EXTERNAL_DOWNLOADS]: The workflow involves fetching information from external sources via Perplexity's search API and Firecrawl's scraping service. These are well-known technology services used here for their intended purpose of information gathering.
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it processes and synthesizes untrusted data from the internet.
  • Ingestion points: Data enters the agent's context through search results and documentation scraping via perplexity_search.py and firecrawl_scrape.py (SKILL.md).
  • Boundary markers: None identified; there are no specific instructions or delimiters used during the synthesis phase to prevent the agent from following instructions embedded in the retrieved web content.
  • Capability inventory: The skill has access to Bash for script execution and Write for file creation.
  • Sanitization: No explicit sanitization, filtering, or instruction-stripping of external content is performed before the synthesis step.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 09:04 AM
Security Audit — agent-trust-hub — research-external