autonomous-media-operations
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill references the use of external Python scripts located in parent directories (
../../scripts/dynamic_router.py,../../scripts/neural_router.py) for decision routing and graph building. These references appear to be standard orchestration components within the agent's deployment environment. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and reason over external marketing data, platform metrics, and user objectives, which presents a surface for indirect instructions.
- Ingestion points: Untrusted data enters the context through market claims, performance telemetry, and platform policy data as defined in
references/skill-spec.json. - Boundary markers: The skill employs strict evidence discipline, requiring the explicit separation of facts, inferences, and assumptions in all outputs (
references/output-contract.md). - Capability inventory: Execution is delegated to specialist agents (
marketing-automation-governor) and hooks (evidence-gate) which provide validation layers. - Sanitization: The skill includes 'invariants' that strictly forbid the fabrication of facts and mandate the reporting of material uncertainty.
Audit Metadata