The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a multi-agent orchestration pattern where a Coordinator ingests data from multiple subagents. This creates a surface for indirect prompt injection if subagents process untrusted external data (e.g., web research, third-party code).
Ingestion points: The Coordinator receives status updates and final reports via send_message from subagents, and ingests deliverables from workspace artifacts (e.g., staged files in the staging/ directory).
Boundary markers (absent): The skill documentation advises on Explicit Output Contracts for subagents but does not explicitly mandate the use of strict delimiters or ignore instructions warnings within the Coordinator synthesis logic.
Capability inventory: The Coordinator utilizes invoke_subagent, manage_subagents, and send_message to manage the lifecycle of workers.
Sanitization (absent): No specific sanitization or filtering of subagent-provided content is defined in the operational protocols.
[SAFE]: The skill references a catalog at https://skills.danicat.dev, which is a legitimate vendor resource owned by the author (danicat).

parallel