nyc-hpd
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the NYC Open Data API (data.cityofnewyork.us) to retrieve building violations and registrations. This external content is then processed and rendered by the agent, creating a surface where malicious instructions embedded in the building data could potentially influence agent behavior.
- Ingestion points: Data is retrieved from various datasets on data.cityofnewyork.us (wvxf-dwi5, csn4-vhvf, ygpa-z7cr, tesw-yqqr, feu5-w2e2) using the WebFetch tool.
- Boundary markers: The results are formatted using Markdown headers and tables, but there are no specific instructions or delimiters used to ensure the agent ignores any command-like text that might be present in violation descriptions or owner names.
- Capability inventory: The skill is granted access to WebFetch, Write, Read, and Bash tools, which provides a high-capability environment for any potential injection to exploit.
- Sanitization: The instructions do not describe any sanitization, filtering, or validation of the retrieved data before it is presented or potentially written to files.
Audit Metadata