The Agent Skills Directory

[Content Sanitization Protocol]: The skill implements a comprehensive protocol to ensure sensitive data is not leaked when generating work items. It specifically instructs the agent to replace absolute file paths with workspace-relative references and to remove internal state JSON, which helps maintain data privacy and prevents the exposure of internal system structures.
[Human-in-the-Loop Governance]: The 'Autonomy-Tier Enumeration' section defines how the agent interacts with external tools. By setting 'supervised' as the default tier, the skill ensures that batches of work items are presented for user review and approval before the agent proceeds with automated creation, maintaining appropriate human oversight.
[AI Transparency Measures]: The skill requires the inclusion of a verbatim disclaimer block in all generated artifacts. This informs the end-user that the content was AI-assisted and requires manual validation by a qualified reviewer, which is a best practice for AI safety and accountability.
[Data Handoff Conventions]: The use of standardized ID prefixes and structured templates for ADO and GitHub ensures that cross-planner references are traceable and consistent, reducing the risk of data collision or misinterpretation during complex project planning workflows.

backlog-templates