orchestration

Fail

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PRIVILEGE_ESCALATION]: The skill explicitly directs the agent to attempt to bypass sandbox security by requesting elevated permissions if a runtime access error occurs. Evidence: 'If it fails with runtime_access_denied, your sandbox blocked the connection: re-run it with escalated permissions' in SKILL.md.
  • [COMMAND_EXECUTION]: The skill relies on executing system binaries (orca, orca-ide, orca-dev) to perform its core functions. Evidence: instructions to resolve the executable once and reuse it for all orchestration commands.
  • [INDIRECT_PROMPT_INJECTION]: The agent is instructed to load and act upon instructions generated dynamically by a tool, creating a vulnerability surface. 1. Ingestion points: Output from 'ORCA skills get orchestration' and specific reference files requested via the --reference flag. 2. Boundary markers: Absent. 3. Capability inventory: Execution of arbitrary orchestration and shell commands via the Orca CLI. 4. Sanitization: Absent; tool output is followed as a trusted guide.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 9, 2026, 02:32 AM