The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user artifacts (such as code, writing, or designs) and passes them to subordinate agents (Orca members) for critique. This creates a surface where instructions embedded within the artifact could influence the agents' behavior or scoring decisions.
Ingestion points: User-supplied artifacts (files, text, screenshots) provided in the initial prompt or workspace.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when passing artifacts to council members.
Capability inventory: The skill orchestrates other agents using the ORCA CLI and performs file system operations (read/write) in the .council/ directory.
Sanitization: There is no mentioned sanitization or filtering of the artifact content before it is processed by the council members.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via the ORCA CLI to manage its workflow and orchestrate other agents.
Evidence: Execution of ORCA status --json, ORCA orchestration worker-start, and worker-list commands to manage the state and execution of the council members.

council