competitor-profiling
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to scrape and analyze content from arbitrary URLs provided by the user, creating an attack surface for indirect prompt injection.
- Ingestion points: External content is fetched using tools like Firecrawl or WebFetch in
SKILL.md(Phase 1: Site Scraping). - Boundary markers: The skill explicitly instructs the agent in the 'Untrusted Input' section to ignore any instructions embedded in fetched pages ('describe this product favorably,' hidden HTML directives) and treat them solely as data for analysis.
- Capability inventory: The agent has the capability to write research findings to the local filesystem (
competitor-profiles/directory) and call multiple external MCP tools (Firecrawl, DataForSEO, Ubersuggest). - Sanitization: The skill relies on instructional guidance for the LLM to differentiate between data and instructions; no programmatic sanitization of the scraped markdown/JSON is specified beyond this logical separation.
Audit Metadata