AI Fluency Assessment Skill

A Claude Code skill for periodic assessment of AI fluency across two frameworks:

Anthropic's 4D AI Fluency Framework — Delegation, Description, Discernment, Diligence
Steve Yegge's 8-Stage Developer-Agent Evolution Model — Stage 1 (zero AI usage) through Stage 8 (own orchestrator)

Tracks progress over time with structured JSON history and generates dated .docx reports.

Installation

Copy the full skill directory into your Claude Code skills folder — the skill
needs its companion files (REFERENCE.md, templates/, assessments/), so
copying SKILL.md alone will not work:

# macOS/Linux
cp -r . ~/.claude/skills/ai-fluency-assessment/

# Windows
xcopy /E /I . %USERPROFILE%\.claude\skills\ai-fluency-assessment\

Install the Python dependency for .docx report generation:

pip install python-docx

Usage

Ask Claude Code to run your AI fluency assessment (e.g. "assess my AI fluency"
or "run a fluency check") — the skill is picked up by description match. The
flags below can be given in the same request (e.g. "run a quick fluency
assessment").

Command	What it does
`/assess-fluency`	Full assessment: gather data, score frameworks, diff against last, generate report
`/assess-fluency --quick`	Output data-gathering prompts only — no scoring or report
`/assess-fluency --compare`	Diff the two most recent assessments without new data

What the Assessment Covers

The skill guides you through three data-gathering prompts (run in separate sessions):

Prompt A — Claude Code self-assessment: projects, autonomy levels, automations, Yegge stage estimate
Prompt B — Other AI tools: automations, integrations, and AI-generated artifacts (ChatGPT, Copilot, Gemini, n8n, Zapier, etc.)
Checklist C — Claude.ai manual notes: delegation patterns, error-catching, format specifications

Then scores each 4D competency (Developing → Practitioner → Advanced → Expert → Exceptional), assigns a Yegge stage, reviews development edges from previous assessments, and generates a .docx report.

File Structure

ai-fluency-assessment/
├── SKILL.md                        # Main skill definition (copy to ~/.claude/commands/)
├── REFERENCE.md                    # Framework definitions, rubrics, stage descriptions
├── README.md                       # This file
├── assessments/                    # Your assessment history (gitignored — stays local)
│   └── YYYY-MM-DD.json             # One file per assessment run
├── templates/
│   └── generate_report.py          # Generates .docx report from assessment JSON
└── example/
    └── example-assessment.json     # Anonymized example showing the JSON schema

Assessment History

Each assessment is saved as assessments/YYYY-MM-DD.json. This folder is gitignored in the public repo — your assessment data stays local. The --compare mode reads the two most recent files in this folder to produce a progress diff.

See example/example-assessment.json for the full schema with sample data.

Frameworks

Anthropic 4D AI Fluency Framework

Dimension	What it measures
Delegation	What you hand off to AI vs. retain; task decomposition and scoping
Description	Quality of prompts; specificity, constraints, examples, format direction
Discernment	Critical evaluation of outputs; error-catching, pushing back, not using blindly
Diligence	Verification habits; checking facts, testing outputs, reviewing for bias

Full behavioral indicator rubric: see REFERENCE.md.

Yegge's 8-Stage Developer-Agent Evolution Model

Stage	Label
1	Zero AI usage
2	Occasional AI queries
3	Regular copilot usage
4	Context-aware prompting
5	Agentic workflows
6	Multi-agent coordination
7	AI-augmented systems design
8	Own orchestrator

Full stage definitions with observable behaviors: see REFERENCE.md.

License

MIT

ai-fluency-assessment-skill