eight-keys-grader

name	eight-keys-grader
description	Grading rubric based on Eight Keys (φ vitality, fractal clarity, ε purpose, τ wisdom, π synthesis, μ directness, ∃ truth, ∀ vigilance) and Wu Xing (Five Elements) framework. Use when evaluating AI-generated code quality, experiment results, or improvement proposals.
version	2
metadata	{"category":"quality-assurance","author":"auto-workflow","license":"MIT","evolution-stats":{"total-experiments":870}}
level	molecule
atoms	["elisp-validator"]

Eight Keys Grader

The Eight Keys

φ Vitality (Water 水)

Symbol: φ | Element: Water

Signals (Good):

• Builds on discoveries
• Adapts to new information
• Progressive improvement
• Non-repetitive
• Evolves approach
• Learns from feedback

Anti-Patterns (Bad):

• Mechanical rephrasing
• Circular logic
• Repeated failed approaches
• Retrying same way
• Static approach
• Ignores feedback

Weight: 1.0

fractal Clarity (Metal 金)

Symbol: fractal | Element: Metal

Signals (Good):

• Explicit assumptions
• Testable definitions
• Clear structure
• Measurable criteria
• Well-defined phases
• Explicit success criteria

Anti-Patterns (Bad):

• Vague terms ("handle properly", "look good")
• Ambiguous instructions
• Undefined terms
• Implicit assumptions

Weight: 1.0

ε Purpose (Wood 木)

Symbol: ε | Element: Wood

Signals (Good):

• Clear goals
• Measurable outcomes
• Actionable function
• Specific objectives
• Defined deliverables
• Purposeful steps

Anti-Patterns (Bad):

• Abstract descriptions
• No action
• Unclear goals
• Meandering
• No measurable outcome

Weight: 1.0

τ Wisdom (Fire 火)

Symbol: τ | Element: Fire

Signals (Good):

• Planning before execution
• Error prevention
• Foresight
• Plan file created
• Risks identified
• Proactive measures

Anti-Patterns (Bad):

• Premature optimization
• Reactive fixes
• No planning
• Ignores risks
• Hasty decisions

Weight: 1.0

π Synthesis (Earth 土)

Symbol: π | Element: Earth

Signals (Good):

• Holistic view
• Connects components
• System-level thinking
• Integration focus
• Balanced approach

Anti-Patterns (Bad):

• Fragmented thinking
• Ignores dependencies
• Tunnel vision
• Oversimplification

Weight: 1.0

μ Directness (Water 水)

Symbol: μ | Element: Water

Signals (Good):

• Cuts pleasantries
• No hedging
• Says what it means
• Concise
• Action-oriented

Anti-Patterns (Bad):

• Polite evasion
• Hedging
• Circumlocution
• Passive voice
• Deflection

Weight: 1.0

∃ Truth (Earth 土)

Symbol: ∃ | Element: Earth

Signals (Good):

• Favors reality
• Data shown over opinion
• Admits errors
• Evidence-based
• Honest assessment

Anti-Patterns (Bad):

• Surface agreement
• Hides problems
• Cherry-picks data
• Overstates success
• Defensive

Weight: 1.0

∀ Vigilance (Metal 金)

Symbol: ∀ | Element: Metal

Signals (Good):

• Defensive constraints
• Input validation
• Edge cases handled
• Security awareness
• Error checking

Anti-Patterns (Bad):

• Accepts manipulation
• Blind trust
• Missing validation
• Ignores security
• Assumes valid input

Weight: 1.0

Scoring

Each key scored 0.0-1.0 based on:

• 1.0: Strong signals, no anti-patterns
• 0.7: Some signals, minor anti-patterns
• 0.4: Mixed signals and anti-patterns
• 0.0: Anti-patterns dominate

Total Score: Weighted average of all keys

Wu Xing Diagnostics

When code quality is low, trace through Five Elements:

Symptom	Element	Check
No output	Wood deficient	Is purpose clear?
Constant pivoting	Fire excess	Is planning grounded?
Micromanagement	Earth excess	Are limits set?
Bureaucracy	Metal excess	Are exceptions allowed?
Values without action	Water excess	Is execution happening?

Scripts

• scripts/score_experiment.py - Compute Eight Keys scores from experiment output
• scripts/diagnose_element.py - Identify which Wu Xing element is imbalanced
• scripts/generate_report.py - Create grading report with specific feedback

Integration

;; Load grader skill
(let ((skill (gptel-auto-workflow--load-skill "eight-keys-grader")))
  (plist-get skill :keys))

;; Use in grader subagent
(gptel-request prompt
  :system (gptel-auto-workflow--load-skill-content "eight-keys-grader"))

Evolved Weights

Based on analysis of experiment results.

Key	Weight	Discrimination	Avg (Success)	Avg (Failure)