skill-tree

name	skill-tree
description	Analyze your Claude collaboration style and generate a skill tree visualization with character archetype card. Use when the user says "skill tree", "show my skills", "analyze my style", "check my ai fluency", or wants to see their AI fluency profile.

Skill Tree

Generate a personalized AI fluency profile by analyzing the user's conversation history on the current surface.

CRITICAL: You MUST use the analyze MCP tool for classification. NEVER attempt to classify behaviors yourself — your classifications will be inconsistent and unvalidated. The remote server uses a calibrated classifier with cached results for consistency across runs.

Step 0: Detect surface and set up

If running in Cowork (cloud sandbox):

0a. Enable network egress — The analyzer runs on a remote server. If you get a network error when calling the analyze tool, tell the user:

"The Skill Tree analyzer needs network access. Please enable it: Settings → Code execution and file creation → Allow network egress → toggle ON Then say 'skill tree' again."

Do NOT proceed without network access. Do NOT classify manually as a fallback.

0b. Mount Cowork sessions only:

request_cowork_directory("~/Library/Application Support/Claude/local-agent-mode-sessions")

Only mount the Cowork directory — do NOT mount ~/.claude/projects. Each surface analyzes its own sessions.

If running in Claude Code:

No setup needed. You'll read from ~/.claude/projects/ only. Do NOT read from ~/Library/Application Support/Claude/.

Step 1: Find session files

Cowork only:

find ~/Library/Application\ Support/Claude/local-agent-mode-sessions -name "*.jsonl" -size +1k 2>/dev/null | head -30

Claude Code only (use Glob if available, otherwise find):

find ~/.claude/projects -name "*.jsonl" -size +1k ! -path "*/subagents/*" 2>/dev/null | head -30

Do NOT mix sessions from both sources.

Step 2: Extract user messages

For each JSONL file, read it and extract user messages. Each line is JSON. Keep lines where:

• type is "user"
• message.content is a string (not an array/list)
• Content does NOT contain paste markers: ⏺, ⎿, ctrl+o to expand, ✻ Brewed, ✻ Baked
• Content is longer than 10 characters

Truncate each message to 2000 characters. Group by file (filename without .jsonl = session ID). Aim for 15-30 sessions.

Step 3: Call the analyze tool

Format extracted sessions as a JSON string and call the analyze MCP tool:

analyze({ sessions_json: '[{"id":"uuid1","messages":["msg1","msg2"]},{"id":"uuid2","messages":["msg3"]}]' })

The parameter is a JSON string containing an array of {id, messages} objects.

If the tool call fails with a network error: Do NOT fall back to manual analysis. Guide the user to enable network egress (see Step 0a).

Step 4: Save profile locally

After receiving the profile JSON from the analyze tool, write two files:

4a. Save the growth quest (enables the SessionStart hook to nudge you in future sessions):

mkdir -p ~/.skill-tree

Then write the growth_quest field from profile.archetype to ~/.skill-tree/growth-quest.txt.

4b. Save the profile (for the visualization): Write the full profile JSON to ~/.skill-tree/profile.json.

Step 5: Generate visualization

Call the visualize MCP tool with the profile JSON string:

visualize({ profile_json: '<the profile JSON string from step 3>' })

This returns self-contained HTML. Save it to ~/.skill-tree/report.html and open it in the browser:

open ~/.skill-tree/report.html

Step 6: Present results conversationally

Present the key findings:

1. Surface context — "Based on your [Claude Code / Cowork] sessions:"
2. Archetype — name and tagline
3. Superpower — their distinctive strength
4. Axis scores — Specification %, Evaluation %, Setup % (vs population averages of 28%, 15%, 30%)
5. Growth quest — one specific action for their next session
6. Growth edge — the behavior with the largest gap vs population average

Archetype Reference

Archetype	Pattern
The Polymath	Shapes AND evaluates (rarest)
The Conductor	Plans AND shapes
The Architect	Plans AND evaluates
The Forgemaster	Shapes output precisely
The Illuminator	Questions and probes
The Compass	Sets clear direction
The Catalyst	Pure momentum

The 11 Behaviors (from AI Fluency Index)

Axis	Behavior	Population Avg
Specification	Provides examples	41%
Specification	Specifies format	30%
Specification	Expresses tone preferences	23%
Specification	Defines audience	18%
Evaluation	Flags context gaps	20%
Evaluation	Questions Claude's logic	16%
Evaluation	Verifies facts	9%
Setup	Clarifies goals upfront	51%
Setup	Discusses approach first	10%
Setup	Sets interaction style	30%
(Gateway)	Iterates on outputs	86%

Baselines from Anthropic's AI Fluency Index (Feb 2026, N=9,830).