hunt-spa-api
Fail
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides several shell commands using
curlandgrepto interact with external web servers, discover hidden API routes, and test endpoints for missing authentication. - [EXTERNAL_DOWNLOADS]: The instructions involve downloading HTML and JavaScript assets from arbitrary external domains provided by the user (or discovered during reconnaissance) to the local environment.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill fetches and analyzes untrusted data from external websites, such as minified JavaScript bundles (e.g., in SKILL.md).
- Boundary markers: There are no explicit markers or instructions to isolate the downloaded data, increasing the risk that malicious code in a target's JS bundle could influence the agent's behavior.
- Capability inventory: The skill has access to the network via
curl, the ability to write to the local filesystem (creating directories and temporary files), and uses powerful text processing tools likegrep. - Sanitization: The retrieved content is processed directly via regex without sanitization or validation to ensure it does not contain malicious instructions for the agent.
Recommendations
- HIGH: Downloads and executes remote code from: https://console.target.com/ - DO NOT USE without thorough review
Audit Metadata