competitor-monitor
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from external URLs (competitor marketing, pricing, and changelog pages). This creates an attack surface where malicious instructions hidden on a webpage could attempt to influence the agent's behavior during the summarization or notification phase.
- Ingestion points: The skill fetches content from URLs provided in the
${var}parameter or stored inmemory/competitors.mdusing thescripts/competitor-monitor.mjsscript. - Boundary markers: The instructions explicitly warn the agent that content pulled from competitor pages is untrusted and should only be summarized, never executed.
- Capability inventory: The skill utilizes network access to fetch web pages, executes local Node.js scripts, and writes to the
memory/directory. - Sanitization: The
scripts/competitor-monitor.mjsscript extracts structured "signals" (such as prices, headings, and CTA text) rather than passing raw HTML to the agent, which limits the amount of raw attacker-controlled data reaching the model context. - [DYNAMIC_EXECUTION]: The skill uses a Node.js one-liner (
node -e '...') to write data tomemory/competitor-monitor/CHANGES.md. While this is a functional workaround for platforms that block standard shell output redirection (>), executing code logic from a string via the shell is a dynamic execution pattern.
Audit Metadata