The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly directs the agent to "re-run it with escalated permissions" if a runtime_access_denied error occurs, instructing the agent to bypass security sandboxes or access controls.
[DYNAMIC_EXECUTION]: The skill uses dynamic logic to resolve the executable path for the ORCA tool, checking environment variables such as ORCA_CLI_COMMAND and ORCA_DEV_REPO_ROOT. This execution from computed paths can lead to the execution of unintended binaries if the environment is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to load its usage reference by running ORCA skills get orchestration, creating a vulnerability where the output of the CLI tool can inject or override instructions in the agent's context.
Ingestion points: Output from the ORCA skills get orchestration command.
Boundary markers: None; the agent is instructed to read the reference as a guide for its next actions.
Capability inventory: Execution of shell commands (ORCA open, ORCA skills get) and environment variable resolution.
Sanitization: None mentioned; the agent is expected to trust and follow the guide output.
[COMMAND_EXECUTION]: The skill relies on the execution of dynamically resolved shell commands as its primary coordination mechanism.

orchestration