pp-nypl-digital-collections

Pass

Audited by Gen Agent Trust Hub on Jul 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install external tools using npx and go install. These downloads originate from the vendor's own repositories on NPM (@mvanhorn/printing-press-library) and GitHub (github.com/mvanhorn/printing-press-library).
  • [DATA_EXFILTRATION]: The CLI includes a --deliver webhook:<url> feature. This allows the command output to be automatically POSTed to an arbitrary external URL. While presented as a feature for routing data, it provides a functional path for automated data exfiltration.
  • [COMMAND_EXECUTION]: The skill is primarily a wrapper around a binary (nypl-digital-collections-pp-cli), requiring the execution of various shell commands to retrieve and manage library collection data.
  • [PROMPT_INJECTION]: The skill processes data from the NYPL Digital Collections API, which is an untrusted external source. This introduces a surface for indirect prompt injection if the API returns malicious metadata that the agent later interprets as instructions.
  • Ingestion points: Data enters the context via items search-digital, collections list, and other retrieval commands in SKILL.md.
  • Boundary markers: No specific delimiters or "ignore instructions" markers are used when processing API responses.
  • Capability inventory: The agent has access to Bash, file system writes (via --deliver file), and network access (via --deliver webhook).
  • Sanitization: No explicit sanitization of API data is performed before it is presented to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 19, 2026, 05:44 PM
Security Audit — agent-trust-hub — pp-nypl-digital-collections