robot-update-package

Create structured robot update packages from child dialogue bundles by processing summaries into JSON format with inputs, advice, proposed actions, risks, and open questions. The output defaults to pending review status, preventing automatic application and ensuring human oversight of all changes.

A review-first workflow routes all changes through human validation before application, unlike automatic updates. This Skill enforces pending status on output packages, requiring explicit approval and maintaining audit trails for compliance and traceability.

Yes. Import child dialogue summaries and logs; the tool outputs JSON packages under exports/robot-update-packages with structured fields and pending review status, creating an audit-ready format designed for human-in-the-loop validation without exposing raw dialogue content.

This Skill enforces a non-automatic application policy by setting review.status to pending on all generated packages. Output contains only summarized content in structured fields, requiring manual review and explicit approval before any changes take effect.

Provide child dialogue summaries and logs from imports/child/. The Skill processes these into template-based packages with structured fields: inputs, u1_advice, proposed_actions, risks, and open_questions, optionally augmented with memory context for richer review material.

Structured packages isolate actionable information—suggested actions, risks, and questions—from raw dialogue, reducing review burden while maintaining safety. Template-based formatting ensures consistency, traceability, and compliance-ready audit trails across all generated updates.