The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown documentation and YAML metadata. It does not ship with any executable scripts, binaries, or platform-level command injections. It does not request access to sensitive file paths or perform network operations.
[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for processing external Excalidraw files, which are a potential vector for indirect prompt injection if they contain malicious text elements.
Ingestion points: Files ending in .excalidraw or .excalidraw.json as specified in SKILL.md.
Boundary markers: The 'Iron Law' of delegation enforces a strict architectural boundary by requiring all diagram processing to occur within isolated subagents, protecting the main agent's context.
Capability inventory: The skill uses standard agent capabilities to read and write files and delegate tasks to subagents.
Sanitization: The skill explicitly instructs subagents to return only text-only summaries and to never return raw JSON elements to the main agent, providing a natural filter against embedded malicious instructions.

excalidraw