What problem does it solve?
This Skill prevents a PoV from becoming “fake documentation” by forcing you to attach a real feedback loop that either includes lightweight human-in-the-loop (HITL) approval signals or uses an objective, quantified metric.
Core Features & Use Cases
- HITL or objective metric selection: Choose lightweight human approval/rejection/edit capture for consequential actions, or define an objective environment signal for advisory outputs (or both for Tier-1/2).
- Operationalizes feedback collection: Specify how frequently feedback is captured/aggregated and how it is reviewed.
- Defines pivot behavior with quantified thresholds: Set a windowed threshold that triggers a PoV re-scope via
kata-pov-scope-define, distinct from discontinuation.
- Persists feedback for downstream use: Writes
feedback.md to the required docs/{context}/agents-pov/{agent}/ location with cross-links.
Quick Start
Ask an agent to read your pov.md, decide whether you need lightweight HITL, an objective metric, or both, and then generate the required feedback.md at docs/{context}/agents-pov/{agent}/feedback.md.