research-the-market
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to perform web searches and ingest content from untrusted external sources, which creates a potential surface for indirect prompt injection.
- Ingestion points: External web content from subreddits, Hacker News threads, GitHub issues, and various online forums (as described in the 'The five things to find' section).
- Boundary markers: None identified; the instructions do not include specific delimiters or warnings to ignore instructions that might be embedded in the retrieved research data.
- Capability inventory: The skill involves reading local files (
founder-brief.md) and writing research results to a local file (market-research.md). - Sanitization: No explicit sanitization or filtering logic is provided to handle potentially malicious instructions within the data retrieved from the web.
- [COMMAND_EXECUTION]: The skill utilizes the agent's internal browsing capabilities to perform searches. While this involves interacting with external services, it is constrained to the primary purpose of market research and does not include arbitrary shell command execution.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads a project brief to inform its search queries. While this involves taking local context and using it in web searches, the behavior is transparently defined and aligned with the skill's stated purpose of research.
Audit Metadata