The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses tools like create_file and str_replace to manage document drafts and apply iterative edits. These operations are restricted to the local workspace and are necessary for the skill's primary purpose.
[EXTERNAL_DOWNLOADS]: The workflow describes fetching content from external services including Google Drive, SharePoint, Slack, and Microsoft Teams. While these are well-known services, the skill relies on these external inputs to build document context.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) due to its core functionality of processing external data.
Ingestion points: The agent ingest data from external files, shared document links, and messaging platform threads (Slack/Teams).
Boundary markers: There are no explicit boundary markers or instructions to the LLM to ignore potentially malicious instructions embedded within the imported context.
Capability inventory: The agent possesses file-writing capabilities (create_file, str_replace) which could be manipulated if an ingested document contains malicious instructions.
Sanitization: No sanitization or validation of the external content is performed before the data is interpolated into the agent's context.

doc-coauthoring