salary-market-scanner
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Orchestrates multiple sub-agents using shell commands and parallel execution, coordinating results via temporary files in the /tmp directory.
- [EXTERNAL_DOWNLOADS]: Requires the installation of the @tiny-fish/cli tool from npm, which is provided by the skill's author for interacting with the TinyFish platform.
- [DATA_EXFILTRATION]: Directs agents to perform network requests to external domains including google.com, linkedin.com, indeed.com, and levels.fyi to retrieve market data.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating user-supplied role and location variables directly into natural language instructions for sub-agents.
- Ingestion points: User-provided variables ({ROLE}, {LOCATION}) in SKILL.md.
- Boundary markers: None present; user inputs are inserted directly into prompt strings without delimiters.
- Capability inventory: Sub-agents possess network access and the ability to navigate and extract data from external websites via the tinyfish agent run command.
- Sanitization: No sanitization, escaping, or validation of user inputs is performed before they are processed by the agents.
Audit Metadata