The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and process data from external, potentially untrusted sources such as Slack messages, Teams threads, and shared cloud documents.
Ingestion points: External content is retrieved via messaging/document integrations and user-provided links or file paths as described in SKILL.md.
Boundary markers: The instructions lack explicit requirements for using delimiters or 'ignore embedded instructions' warnings when the agent processes these external context streams.
Capability inventory: The skill utilizes create_file and str_replace for document management and invokes sub-agents for reader testing, which could be targeted by malicious instructions in the ingested data.
Sanitization: No specific filtering or sanitization procedures are defined for the content retrieved from external integrations.
[DATA_EXFILTRATION]: The skill defines patterns for connecting to and reading from third-party services including Slack, Microsoft Teams, Google Drive, and SharePoint.
The workflow explicitly mandates user confirmation before the agent searches or reads from these external systems, ensuring the user maintains control over data access.
The integration targets are well-known, established productivity services.
[COMMAND_EXECUTION]: The skill employs file system operations (create_file, str_replace) to build and iterate on document scaffolds.
These operations are restricted to the local working directory or artifact space and are driven by the specific document sections agreed upon with the user.

doc-coauthoring