stock-question-refiner
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides a template that interpolates untrusted user input (stock tickers and company names) directly into directory and file path instructions (e.g.,
RESEARCH/STOCK_[ticker]_[company_name]/) for a downstream agent. - Ingestion points: User-provided
tickerandcompany_namecollected during the initial interaction flow (Step 1) and used to generate the research prompt (Step 3). - Boundary markers: There are no delimiters or warnings in the instructions to prevent the agent from interpreting special characters within the user-provided strings.
- Capability inventory: The skill instructions specify a multi-file write operation (
01_Business_Foundation.md, etc.) based on the generated path, which involves file system interaction. - Sanitization: The skill lacks instructions to sanitize or validate the user-provided strings for path traversal characters (like
..or/) or shell metacharacters, which could lead to files being written outside the intended directory if the downstream agent is not independently hardened.
Audit Metadata