The Agent Skills Directory

[SAFE]: No malicious patterns, unauthorized data exfiltration, or obfuscation techniques were identified. The skill correctly implements its intended functionality for document collaboration using vendor-specific tools and domains.
[PROMPT_INJECTION]: The skill handles document content that may contain untrusted data, presenting a potential surface for indirect prompt injection. This risk is effectively addressed by instructions to use structured API interactions and clear boundary markers for rich content components.
Ingestion points: Document content and comments are read through the read_docs tool as described in SKILL.md and the reference matrix.
Boundary markers: The skill employs specific italicized slot markers (e.g., _[mention: <user-ref>]_) and structured JSON delta_format objects to provide distinct boundaries for different data types.
Capability inventory: Capabilities are limited to document-specific operations via create_doc, update_doc, delete_block, and add_comment tools.
Sanitization: The skill utilizes the platform's structured API for all updates, avoiding raw string interpolation into executable contexts or shell commands.

docs-collaborator