skill-usage-instrumenter

What problem does it solve?

This Skill replaces assumptions with evidence by designing how a skill library records, reviews, and interprets real invocation patterns, so pruning and trigger tuning are based on signal instead of gut feel.

Core Features & Use Cases

• Invocation measurement: Defines what to capture for skill fires, including auto versus explicit activation, coarse task class, and day-level timing.
• Misfire and completion analysis: Tracks wrong-fire corrections, abandonment hints, and never-fired detection over a stated window.
• Privacy-preserving design: Enforces strict minimization so only skill names, enums, counts, and dates are captured, never prompts, responses, identifiers, or live repo details.
• Decision thresholds: Connects usage patterns to actions such as trigger re-review, discriminating trigger fixes, or evidence handoff to a deprecation planner.
• Safe-exception handling: Preserves rare-but-critical skills from being condemned by low usage alone, especially safety-net or incident-response skills.

Quick Start

Ask for a usage instrumentation design for a skill library that identifies which skills fire, which misfire, and which have never fired without collecting any prompt or user-identifying content.