The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes a workflow that incorporates untrusted data from external environments, creating a surface for indirect prompt injection.
Ingestion points: Data is gathered via integrations and MCP servers from Slack threads, Teams channels, Google Drive, and SharePoint documents, as well as unstructured 'info dumps' provided by the user in SKILL.md.
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions when processing or drafting from the ingested external context.
Capability inventory: The agent is instructed to use create_file to generate document artifacts and str_replace to perform surgical edits on existing files. It also has the capability to invoke sub-agents to test document clarity.
Sanitization: There are no instructions for sanitizing or validating external content before it is used for brainstorming or drafting purposes.

doc-coauthoring