marketing-competitor-profiling

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites, creating a surface for indirect prompt injection attacks where malicious instructions could be embedded in competitor pages.
  • Ingestion points: The skill uses the firecrawl_scrape and firecrawl_search tools in SKILL.md and references/tool-reference.md to retrieve raw markdown and text from user-specified competitor URLs and third-party review sites (G2, Capterra).
  • Boundary markers: The skill includes a mandatory security principle in SKILL.md under 'Core Principles #5 (Untrusted Input)' which explicitly labels external data as untrusted and instructs the agent to ignore embedded AI directives.
  • Capability inventory: Across SKILL.md and references/tool-reference.md, the skill utilizes network-enabled MCP tools (Firecrawl and DataForSEO) and performs extensive local filesystem writes to the competitor-profiles/ directory.
  • Sanitization: The instructions require the agent to analyze the content strictly for data extraction and to report any detected injection attempts in the final output rather than obeying them.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 01:07 PM
Security Audit — agent-trust-hub — marketing-competitor-profiling