orchestration
Fail
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill explicitly instructs the agent to "re-run it with escalated permissions" if a
runtime_access_deniederror is encountered, which is a direct instruction to bypass security sandbox boundaries. - [COMMAND_EXECUTION]: The instructions direct the agent to resolve and execute various system-level binaries (orca, orca-dev, orca-ide) to manage terminal control and agent orchestration.
- [DYNAMIC_EXECUTION]: The skill implements a pattern where the full orchestration reference is loaded at runtime via 'ORCA skills get orchestration', allowing for the injection of new instructions into the agent's context after the initial security review.
- [INDIRECT_PROMPT_INJECTION]: The dynamic instruction-loading mechanism creates an attack surface. Evidence: 1. Ingestion point: Output from 'ORCA skills get orchestration' (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Subprocess calls via Orca, terminal control, and shell commands (SKILL.md). 4. Sanitization: Absent.
Recommendations
- AI detected serious security threats
Audit Metadata