The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation suggests running npx agentic-flow commands to initialize swarms and manage agent life cycles.
[EXTERNAL_DOWNLOADS]: The skill requires the agentic-flow package, which is fetched from the npm registry when executed via npx.
[PROMPT_INJECTION]: The orchestration system processes high-level goals and tasks, creating a surface for indirect prompt injection.
Ingestion points: File SKILL.md contains parameters like the --task CLI argument and the goal field in code snippets that ingest natural language instructions.
Boundary markers: No boundary markers or specific 'ignore' instructions are present in the examples to isolate task input from system commands.
Capability inventory: The skill manages agent spawning, task distribution (parallel/pipeline), and shared memory access.
Sanitization: No sanitization, validation, or escaping of external task strings is demonstrated in the provided orchestration logic.

Swarm Orchestration