The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified in the skill's instructions or logic. All operations align with the stated goal of documentation remediation and generation.
[SAFE]: The skill implements clear safety boundaries, such as 'Docstring-Specific Rules' which explicitly prohibit the agent from modifying surrounding code logic when editing comments, maintaining the integrity of the codebase.
[SAFE]: File system operations and shell commands (like cat and grep) are used appropriately for reading configuration and searching text within the project scope, with no evidence of unauthorized data access.
[PROMPT_INJECTION]: The skill processes markdown and source code as data, which is a common surface for indirect prompt injection. This risk is mitigated by strict editing scopes and rules requiring human approval for major changes, ensuring the agent remains under user control.
Ingestion points: Processes local markdown files (**/*.md) and docstrings within code files via the Read tool.
Boundary markers: Uses specific instructions ('ONLY modify docstring/comment text') to restrict modification scope.
Capability inventory: Uses Write and Edit tools to update documentation.
Sanitization: Incorporates human-in-the-loop approval and automated quality gates to verify changes.

doc-generator