skill-logger

license	Apache-2.0
name	skill-logger
description	Logs and scores skill usage quality, tracking output effectiveness, user satisfaction signals, and improvement opportunities. Expert in skill analytics, quality metrics, feedback loops, and continuous improvement. Activate on "skill logging", "skill quality", "skill analytics", "skill scoring", "skill performance", "skill metrics", "track skill usage", "skill improvement". NOT for creating skills (use agent-creator), skill documentation (use skill-coach), or runtime debugging (use debugger skills).
allowed-tools	Read,Write,Edit,Bash,Grep,Glob
category	Productivity & Meta
tags	["logging","analytics","metrics","quality","improvement"]
pairs-with	[{"skill":"automatic-stateful-prompt-improver","reason":"Data for prompt optimization"},{"skill":"skill-coach","reason":"Quality tracking feeds coaching"}]

Skill Logger

Track, measure, and improve skill quality through systematic logging and scoring.

Decision Points

Skill Category → Logging Signals Priority

Skill Category Analysis:
├─ Code Generation Skills
│  ├─ IF output type = "code" → Priority: syntax_correctness, test_pass_rate, user_edits
│  ├─ IF includes tool calls → Track: tool_success_rate, retry_count
│  └─ ELSE → Standard completion signals
│
├─ Analysis/Advisory Skills  
│  ├─ IF output is recommendations → Priority: follow_up_rate, acceptance_rate
│  ├─ IF research-heavy → Track: source_quality, comprehensiveness
│  └─ ELSE → Focus on user_satisfaction, edit_ratio
│
├─ Creative/Content Skills
│  ├─ IF artistic output → Priority: user_acceptance, revision_requests
│  ├─ IF writing/documentation → Track: readability_score, user_edits
│  └─ ELSE → Standard quality metrics
│
└─ Meta/System Skills
   ├─ IF affects other skills → Priority: downstream_impact, system_health
   ├─ IF automation focused → Track: execution_success, error_recovery
   └─ ELSE → Completion rate, efficiency metrics

Quality Signal Collection Strategy

Signal Availability Decision Tree:
├─ Real-time signals available?
│  ├─ YES → Collect: completion_rate, token_efficiency, tool_success
│  └─ NO → Skip to delayed collection
│
├─ User interaction possible?
│  ├─ YES → Request: thumbs_up/down, edit_ratio measurement
│  └─ NO → Infer from: retry_requests, follow_up_questions
│
├─ Output testable?
│  ├─ Code → Run: syntax_check, basic_execution
│  ├─ Data → Validate: format_compliance, completeness
│  └─ Text → Check: length_appropriateness, structure
│
└─ Delayed validation available?
   ├─ YES → Schedule: outcome_tracking, revert_detection
   └─ NO → Mark as: immediate_signals_only

Scoring Model Selection

Based on skill usage context:
IF high_stakes_usage (production, important decisions):
   → Use strict scoring: require 90+ for "good", weight errors heavily
   
ELIF experimental_usage (testing, learning):
   → Use lenient scoring: 70+ acceptable, focus on learning signals
   
ELIF routine_usage (daily workflow):
   → Use balanced scoring: standard thresholds, efficiency emphasis
   
ELSE (unknown context):
   → Default to balanced scoring with conservative error handling

For the detailed multi-dimensional scoring model with specific component weights and calculation formulas, see references/scoring-rubric.md.

Failure Modes

Token Inflation Anti-Pattern

Symptoms: Skill produces unnecessarily verbose outputs, token usage 2x+ expected baseline Detection: IF tokens_output > baseline_tokens * 2.0 AND user_edit_ratio > 0.6 Diagnosis: Skill optimizing for completeness over conciseness Fix: Add explicit length constraints, example-based training on concise outputs