paw-wbc-agent-discovery
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection because its core function involves querying and crawling the public web via search tools. Malicious content on external websites could attempt to influence the agent's research summaries or downstream hook generation.
- Ingestion points: The agent ingests untrusted data from the web through tools like
web_search_exaandcrawling_exa, as well as processing user-provided documents. - Boundary markers: The skill does not define explicit delimiters or instructions to treat external research data as untrusted, though it emphasizes 'ruthless filtering' during the compression phase.
- Capability inventory: The agent can write files to the local project directory (e.g.,
research-context.md,brief.md) and perform further network searches. - Sanitization: There is no formal sanitization process for external content beyond the agent's natural language processing and summary generation.
- [EXTERNAL_DOWNLOADS]: The skill uses external MCP tools provided by Exa for web searching and content extraction. This is consistent with its stated purpose of webinar research and uses well-known specialized services for data retrieval.
Audit Metadata