public-relations
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources to identify trending topics, creating a potential surface for indirect prompt injection if an attacker-controlled post were processed by the agent.
- Ingestion points: RSS and API feeds for Google News, Hacker News, and Reddit as defined in references/newsjacking.md.
- Boundary markers: SKILL.md contains specific instructions requiring human review and approval of all recipients and content before communication.
- Capability inventory: The agent drafts pitches and uses browser tools for journalist research.
- Sanitization: Reliance on human-in-the-loop verification is the primary control mentioned.
- [COMMAND_EXECUTION]: The skill provides utility command snippets for data retrieval and tool installation.
- Evidence: references/newsjacking.md includes curl commands targeting well-known services like Google and Reddit, and recommends installing the jq utility via Homebrew.
Audit Metadata