The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses hive and tmux CLI tools to perform legitimate administrative actions, such as spawning agent peers (hive gang spawn-peer), renaming terminal windows for organization (tmux rename-window), and checking team status (hive team). These operations are consistent with the orchestrator persona.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and processes data (verdicts and status signals) from other agents (validators and skeptics). This is inherent to the multi-agent collaboration workflow described.
Ingestion points: Reads status signals from <gang>.skeptic and integrated validation results.
Boundary markers: No explicit delimiters or sanitization instructions are provided for external agent input.
Capability inventory: The skill can execute commands via hive and write to the local filesystem using the Edit tool.
Sanitization: Not explicitly implemented; however, the logic restricts actions to a specific protocol (flipping status bits or requesting human intervention).

gang-orch