orchestration
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to interact with the host system by executing the
orcacommand-line utility (includingorca-ideandorca-devvariants) to perform agent orchestration tasks. - [INDIRECT_PROMPT_INJECTION]: The system is instructed to load a "version-matched guide" directly from the output of a shell command (
ORCA skills get orchestration). This pattern represents an attack surface for indirect prompt injection, as the agent is directed to follow instructions generated dynamically by a local executable. - Ingestion points: Output of
ORCA skills get orchestrationin SKILL.md. - Boundary markers: None specified for the injected content.
- Capability inventory: Shell execution via
orcaCLI, state monitoring, and process management. - Sanitization: No sanitization or validation of the guide content is mentioned.
- [PRIVILEGE_ESCALATION]: The documentation contains an explicit instruction to re-run commands with escalated permissions if a
runtime_access_deniederror is returned by the sandbox environment.
Audit Metadata