paw-cra-campaign-orchestration

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Agent tool to orchestrate tasks across specialized sub-agents including paw-cra-agent-designer, paw-cra-agent-video-producer, paw-cra-agent-strategist, and paw-cra-quality-control. This multi-agent coordination is the primary intended function of the orchestration engine and is performed within the scope of the project's creative suite directory.\n- [PROMPT_INJECTION]: As an orchestrator, the skill processes campaign briefs which are untrusted external inputs. This ingestion presents a surface for indirect prompt injection, although the structured processing pipeline mitigates the risk.\n
  • Ingestion points: Campaign brief files (e.g., brief.md) parsed in Stage 01.\n
  • Boundary markers: Deliverables are extracted and classified into structured lists before being passed to sub-agents with focused instruction sets rather than raw brief interpolation.\n
  • Capability inventory: The skill possesses file read/write capabilities within the .pawbytes/ project directory and invokes sub-agents via tool calls.\n
  • Sanitization: Stage 01 performs deliverable classification and dependency analysis, effectively filtering raw input into structured production tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 09:36 AM