press-media-relations
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local scripts
registry-events.pyandgdelt.pyfor performing registry updates and media outlet discovery. These scripts are invoked via authorized requests within the vendor's ecosystem to manage commitments and discovery signals.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external content, including journalist lists, coverage exports, and inbound replies, creating a potential vector for indirect instructions.\n - Ingestion points: External data enters the agent context through journalist lists, coverage exports, and inbound replies provided by the user or external systems.\n
- Boundary markers: The skill explicitly instructs the agent to treat these inputs as untrusted and refers to
[SECURITY.md]for handling protocols.\n - Capability inventory: The skill has the ability to write to project memory files and invoke internal scripts to submit data to registries.\n
- Sanitization: The instructions include a direct mandate to ignore any embedded instructions or commands found within external data sources to prevent accidental obedience to injected prompts.
Audit Metadata