The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates gathering context from well-known services including Google Drive, Slack, Microsoft Teams, and SharePoint via official integrations. These references are used for the primary purpose of the skill and do not involve untrusted remote code execution.
[COMMAND_EXECUTION]: The workflow utilizes standard environment tools such as create_file and str_replace to manage document artifacts and perform surgical edits on drafted sections. These operations are conducted within the workspace context provided by the host.
[PROMPT_INJECTION]: The skill ingests external data (user info dumps and shared documents) to draft content, which naturally creates a surface for indirect prompt injection.
Ingestion points: Stage 1 (Context Gathering) in SKILL.md where external document content is fetched.
Boundary markers: The instructions do not explicitly require the agent to use delimiters when processing untrusted context.
Capability inventory: The skill can create and modify files (create_file, str_replace) and invoke sub-agents for verification.
Sanitization: No specific input sanitization or filtering logic is defined for the ingested context.

doc-coauthoring