The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large amounts of data from potentially untrusted external sources, creating a surface for indirect prompt injection.
Ingestion points: The workflow explicitly prompts the user to provide context through 'info dumping', linking to team channels (Slack, Teams), and shared documents (Google Drive, SharePoint, or other MCP servers) in the 'Context Gathering' stage.
Boundary markers: The instructions do not include specific delimiters or 'ignore embedded instructions' directives when processing these external data streams.
Capability inventory: The skill utilizes file system operations (create_file, str_replace) and the ability to invoke sub-agents for 'Reader Testing'.
Sanitization: There is no evidence of validation or filtering for instructions hidden within the external content being summarized or drafted.
[COMMAND_EXECUTION]: The skill uses platform-specific tools to manage document state.
Evidence: The instructions direct the agent to use create_file for initial scaffolds and str_replace for iterative edits to the documentation artifacts or local markdown files.

doc-coauthoring