competitive-pricing-intel

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests untrusted data from external websites, including competitor pricing pages, blogs, and community forums like Reddit.
  • Ingestion points: The skill uses fetch_webpage in Phase 1A and web_search in Phase 1C to collect external data.
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the fetched content.
  • Capability inventory: The skill has the capability to write files to the local file system (saving reports to the clients/ directory).
  • Sanitization: No validation or filtering is performed on the scraped content before it is processed by the model.
  • [COMMAND_EXECUTION]: The documentation suggests a persistence mechanism by providing a crontab entry that executes a command (python3 run_skill.py). While this is intended for automation, it involves scheduled local command execution.
  • [EXTERNAL_DOWNLOADS]: The skill's primary function involves fetching external data via fetch_webpage and the Web Archive. While these are legitimate actions for the skill's purpose, they involve interaction with arbitrary remote sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 06:02 PM