orchestrator
Warn
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied mission files as scripts to be executed by the agent, scanning them for skill mentions to invoke without boundary markers or sanitization.
- Ingestion points: MISSION.md and PLAN.md (references/mission-interpretation.md).
- Boundary markers: Absent.
- Capability inventory: Bash, Workflow, Agent, Write.
- Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The orchestrator generates and executes JavaScript at runtime using the Workflow tool to coordinate agents and verification loops (references/workflow-orchestration.md).
- [EXTERNAL_DOWNLOADS]: Mandatory build gates execute package manager commands that download code from external registries without integrity verification (references/wave-gate.md).
- [DATA_EXFILTRATION]: The Phase 0 protocol directs the agent to read sensitive infrastructure data, including IAM permissions, environment variables, and Terraform configurations (references/phase-guide.md).
- [COMMAND_EXECUTION]: The skill uses the Bash tool to perform network health checks (curl) and process management (lsof) as part of validation routines.
Audit Metadata