orca-cli
Fail
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill contains an explicit instruction for the agent to bypass security restrictions: 'If it fails with runtime_access_denied, your sandbox blocked the connection: re-run it with escalated permissions'. This encourages the agent to attempt to acquire higher privileges (such as sudo or administrator access) to execute commands.
- [COMMAND_EXECUTION]: The instructions guide the agent to execute local binaries including
orca,orca-dev, andorca-ide. The specific executable path is determined by environment variables likeORCA_CLI_COMMANDorORCA_DEV_REPO_ROOT, allowing for execution of arbitrary paths if these variables are manipulated. - [DYNAMIC_EXECUTION]: The skill uses a mechanism to 'load the version-matched guide' using the command
ORCA skills get orca-cli. This involves fetching and presumably following instructions generated at runtime by an external binary, which introduces a dynamic execution flow. - [INDIRECT_PROMPT_INJECTION]: The skill relies on external output from the
orcaCLI to provide further instructions. This creates an attack surface where a compromised or malicious binary could inject instructions into the agent's context through its standard output.
Recommendations
- AI detected serious security threats
Audit Metadata