news-signal-outreach
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize content from external, untrusted sources including news articles, tweets, and LinkedIn posts. This creates an attack surface where instructions embedded in the external content could potentially manipulate the agent's actions during the extraction and qualification phases.
- Ingestion points: The skill fetches data via URLs or raw text in Step 1 (news_input) and performs company research via web search in Step 2.
- Boundary markers: While the skill uses structured extraction steps, it lacks explicit instructions to the agent to disregard or neutralize instructions found within the processed external text.
- Capability inventory: The skill has access to
web-search,contact-finding, andemail-draftingtools, and it stores business configuration data in local JSON files (clients/<client-name>/config/signal-outreach.json). - Sanitization: There is no evidence of sanitization, filtering, or validation of the data retrieved from external sources before it is processed by the model.
- [EXTERNAL_DOWNLOADS]: The skill uses
web-searchand fetches content from arbitrary external domains to identify and evaluate news signals. Interacting with arbitrary external content carries an inherent risk of encountering malicious data or triggers.
Audit Metadata