The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes scripts such as check-runtime-capabilities.mjs and omo-supervision-reporter.ts to execute local CLI commands (orca, omo, gjc, /bin/ps). these commands are used to manage agent terminals, verify runtime readiness, and monitor process trees for supervision purposes.
[DYNAMIC_EXECUTION]: The skill includes an installation script assets/extensions/install-extensions.ts that provisions a TypeScript-based extension (omo-supervision-reporter.ts) into the local ~/.omo/agent/extensions/ directory. This extension is designed to be loaded by the OMO agent at runtime to report status and heartbeats during task execution.
[INDIRECT_PROMPT_INJECTION]: The skill orchestrates child agents and processes their terminal output, which is treated as untrusted data. It implements several defensive measures to handle this attack surface.
Ingestion points: The skill reads terminal screens, cursor positions, and child worker messages (worker_done, question, escalation) via the Orca CLI.
Boundary markers: It defines and delivers a "SUPERVISION CONTRACT" to workers to establish communication protocols.
Capability inventory: The skill has access to terminal management, task dispatch, local CLI execution, and the ability to write local extension files.
Sanitization: The instructions explicitly mandate safe quoting of terminal arguments and strictly forbid splicing untrusted terminal text directly into commands.

orca-orchestration