The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes specialized tools for file manipulation (write_draft, patch_draft) and quality control (run_quality_checks). These tools are used to maintain manuscript structure and academic rigor within the project scope.
[PROMPT_INJECTION]: The skill demonstrates an indirect injection surface by ingesting data from user-editable files.
Ingestion points: Content is retrieved from draft files and concept.md via the read_draft tool during the writing workflow (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or instructions to ignore potential commands embedded in the drafts.
Capability inventory: The skill has permissions to write to the filesystem and execute internal quality-check hooks.
Sanitization: There is no explicit mention of sanitizing or escaping the content read from files before it is processed by the agent.
[DATA_EXFILTRATION]: The skill manages a local persistence layer for snapshots at drafts/.snapshots/ using an internal DraftSnapshotManager. No external network operations or sensitive data access patterns were identified.

draft-writing