The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates fetching document content and context from well-known technology services and cloud platforms, including Google Drive, Slack, Microsoft Teams, and SharePoint, as described in Stage 1 of the SKILL.md file.
[PROMPT_INJECTION]: The workflow is susceptible to indirect prompt injection (Category 8) due to its core function of processing untrusted external data within an agentic context. • Ingestion points: Untrusted data enters the context through shared document links, file uploads, and content retrieved from messaging and storage integrations (Slack, Teams, Google Drive, SharePoint). • Boundary markers: The instructions lack explicit requirements for the use of boundary markers or instructions to ignore embedded commands within the ingested text. • Capability inventory: The skill utilizes capabilities for file creation (create_file), file modification (str_replace), and the invocation of sub-agents for testing purposes, which could be exploited if malicious instructions are processed. • Sanitization: There are no specific instructions for sanitizing, filtering, or validating the content retrieved from external integrations before it is used to influence the agent's drafting or testing behavior.

doc-coauthoring