competitor-monitor
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs provided in the watch list, creating a surface for indirect prompt injection. An adversarial competitor page could attempt to influence the agent's behavior during the summarization or logging phase.\n
- Ingestion points: External web pages fetched via
scripts/competitor-monitor.mjs.\n - Boundary markers: The skill includes a specific instruction to treat competitor content as untrusted and to summarize rather than execute instructions found within the pages.\n
- Capability inventory: The skill has the capability to execute shell commands, write to the
memory/directory, and send notifications.\n - Sanitization: No programmatic sanitization is implemented; the skill relies on the agent's adherence to instructions to avoid executing external content.\n- [DYNAMIC_EXECUTION]: The skill uses the
node -eflag to execute a fixed JavaScript string that writes content from stdin to the filesystem. This is used to bypass shell redirection (>) restrictions in theread-onlyexecution environment.\n- [COMMAND_EXECUTION]: The skill executes various shell commands (sed,printf,grep,mkdir,ls) and a local Node.js script (scripts/competitor-monitor.mjs) to handle the snapshotting and diffing logic.
Audit Metadata