hackernews-frontpage
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted HTML from Hacker News. If a story title contains malicious instructions, they could influence the agent when it reads the skill output. Ingestion points: script.ts (via browse.goto and browse.html). Boundary markers: None present. Capability inventory: Browser navigation and HTML reading. Sanitization: Simple tag stripping and entity decoding are used.
- [COMMAND_EXECUTION]: The browse-client library runs the git command to determine the project root for state file resolution. Evidence: Found in the defaultStateFile function in _lib/browse-client.ts.
Audit Metadata