The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an aggregation phase where results from multiple sub-workers are read and consolidated. This represents a surface where worker output enters the lead agent's context. 1. Ingestion points: Terminal results and worker reports are read in Phase C. 2. Boundary markers: Absent; the instructions rely on structured report statuses but do not mandate delimiters. 3. Capability inventory: Orchestrates sub-agents, writes to scoped directories or worktrees, and coordinates multi-agent workflows. 4. Sanitization: No sanitization or validation of worker-provided data is specified before aggregation.
[COMMAND_EXECUTION]: The skill utilizes platform tools like 'herdr' and 'pi' to manage agent lifecycles. These are utilized as intended for the orchestration and delegation tasks described in the skill's purpose.

swarm