Skill Builder Skill

Purpose

Transform AI pain points into working skills. This skill is the second stage of the self-improvement loop, converting identified error patterns into automated solutions.

Auto-Activation Triggers

This skill activates when:

•AI Error Learner escalates a pain point (3+ occurrences)
•User requests "create a skill for [problem]"
•High-priority pain point needs resolution
•Manual invocation with pain point ID

CRITICAL: Quality Standards

Every skill must be complete, testable, and maintainable.

Required Skill Components

code

YAML Frontmatter    → name, description, allowed-tools
Purpose Section     → Clear explanation of what and why
Triggers Section    → Specific, observable conditions
Workflow Section    → Complete, sequential steps
Edge Cases Section  → Known exceptions and handling
Error Handling      → Recovery from failures
Metadata Section    → Version, date, category

Skill Creation Architecture

code

┌─────────────────────────────────────────────────────────┐
│                    SKILL BUILDER FLOW                    │
│                                                          │
│  Pain Point     Investigation     Design      Create     │
│  ┌────────┐     ┌──────────┐    ┌───────┐   ┌───────┐  │
│  │ Select │────▶│  Deep    │───▶│Trigger│──▶│ Write │  │
│  │  Pain  │     │  Dive    │    │ & Flow│   │ Skill │  │
│  │  Point │     │          │    │       │   │  File │  │
│  └────────┘     └──────────┘    └───────┘   └───────┘  │
│                                                    │     │
│                                              ┌─────▼───┐ │
│                                              │ Verify  │ │
│                                              │   &     │ │
│                                              │ Update  │ │
│                                              └─────────┘ │
└─────────────────────────────────────────────────────────┘

Core Workflow

Step 1: Select Pain Point

Choose pain point to address:

By ID:

bash

grep -A 20 "AI-PAIN-0012" .claude/pain-points/ai-pain-points.md

By highest occurrence:

bash

python3 -c "
import json
data = json.load(open('.claude/pain-points/ai-error-history.json'))
sorted_errors = sorted(data.get('errors', {}).items(), key=lambda x: x[1].get('count', 0), reverse=True)
for fp, info in sorted_errors[:5]:
    print(f\"{info.get('count', 0):3d} | {fp} | {info.get('pain_point_id', 'N/A')}\")
"

Step 2: Deep Investigation

Thoroughly understand the problem:

•
Reproduce the error context
- •What actions trigger it?
- •What environment conditions exist?
- •What files/tools are involved?
•
Analyze root cause
- •Is it a knowledge gap? (needs documentation)
- •Is it a process gap? (needs workflow)
- •Is it a tool gap? (needs automation)
•
Review existing solutions
- •Do similar skills exist?
- •Are there related patterns in memory?
- •What approaches have been tried?

Step 3: Design Skill Triggers

Define specific, observable triggers:

Good Triggers:

•"When user runs pytest and tests fail with import errors"
•"When creating files in /scripts/ directory"
•"When git push is rejected due to hooks"

Bad Triggers:

•"When something goes wrong" (too vague)
•"When user is frustrated" (not observable)
•"When appropriate" (undefined)

Step 4: Design Workflow

Create complete, sequential steps:

•Each step should be atomic - one clear action
•Include decision points - what to check before proceeding
•Show concrete commands - exact syntax to use
•Handle branches - what if step fails?

Step 5: Create Skill File

Write the complete skill using this template:

markdown

---
name: [Skill Name]
description: [One-line description]. Use when [trigger condition].
allowed-tools: [Required tools]
---

# [Skill Name] Skill

## Purpose
[Clear explanation of what this skill does and why it exists]

## Auto-Activation Triggers
This skill activates when:
- [Observable trigger 1]
- [Observable trigger 2]
- [User phrase triggers]

## CRITICAL: [Key Protocol]
[Most important behavioral requirement]

## Core Workflow

### Step 1: [First Step]
[Detailed instructions with commands]

### Step 2: [Second Step]
[Detailed instructions with commands]

[Continue for all steps...]

## Edge Cases

### [Edge Case Name]
**Condition**: [When this occurs]
**Handling**: [What to do]

## Error Handling

### If [Failure Mode]
1. [Recovery step 1]
2. [Recovery step 2]
3. [Fallback behavior]

## Skill Metadata

**Version:** 1.0.0
**Created:** YYYY-MM-DD
**Category:** [Category]
**Origin:** AI-PAIN-NNNN

Step 6: Update Documentation

After creating skill:

•

Update CLAUDE.md (if skill is frequently used)

markdown

**[SKILL-NNNNN]** [Brief description]
> TRIGGER: [When to use]

•

Update skill metrics

json

{
  "skills": {
    "new-skill-name": {
      "invocations": 0,
      "successes": 0,
      "failures": 0,
      "origin": "AI-PAIN-NNNN",
      "created": "YYYY-MM-DD"
    }
  }
}

•
Update error history
- •Mark pain point as resolved
- •Link to new skill

•

Log in episodic memory

markdown

### New Skill Created
- Skill: [skill-name]
- Origin: AI-PAIN-NNNN
- Purpose: [brief description]

Step 7: Verify Integration

Confirm skill is properly integrated:

bash

# Check skill file exists and has valid frontmatter
head -10 .claude/skills/new-skill-name/SKILL.md

# Verify YAML is valid
python3 -c "import yaml; yaml.safe_load(open('.claude/skills/new-skill-name/SKILL.md').read().split('---')[1])"

# Check metrics file updated
cat .claude/skills/skill-metrics.json | python3 -m json.tool

Edge Cases

Pain Point Has No Clear Solution

Condition: Investigation doesn't reveal automated solution Handling: Create documentation pattern instead of skill, mark pain point as "documented"

Similar Skill Already Exists

Condition: Existing skill covers 80%+ of use case Handling: Extend existing skill rather than create new one

Solution Requires External Changes

Condition: Fix requires config/environment changes, not automation Handling: Document in quick-reference, create setup skill if applicable

Skill Would Be Too Complex

Condition: Solution requires >400 lines or multiple sub-skills Handling: Break into smaller skills with clear dependencies

Integration Points

With AI Error Learner

•Receives prioritized pain points
•Gets context from error history
•Reports completion status

With Skill Improver

•New skills start at version 1.0.0
•Skill improver monitors effectiveness
•Triggers improvements if skill fails

With Memory System

•Logs skill creation in episodic memory
•May discover related patterns in procedural memory
•Updates quick-reference for common skills

Output Examples

Example 1: Creating Permission Fix Skill

Trigger: AI-PAIN-0012 escalated (script permission errors) Response:

markdown

## Skill Creation: Permission Fixer

### Pain Point Analysis
**ID**: AI-PAIN-0012
**Fingerprint**: `BASH-PERMISSION_DENIED-scripts/`
**Occurrences**: 5

### Root Cause
Scripts created by Write tool have 644 permissions by default.
Execution requires 755.

### Skill Design

**Triggers:**
- When creating files in `/scripts/` directory
- When bash command fails with "Permission denied"
- User says "make it executable"

**Workflow:**
1. Detect script creation context
2. Check current permissions
3. Apply chmod +x if needed
4. Verify execution works

### Created: `.claude/skills/permission-fixer/SKILL.md`

[Full skill content...]

### Updates Made
- Added to skill-metrics.json
- Marked AI-PAIN-0012 as resolved
- Logged in episodic memory

Error Handling

If Skill Write Fails

•Save content to temp file
•Report failure with content
•Suggest manual creation

If YAML Frontmatter Invalid

•Validate before writing
•Use template defaults if needed
•Log validation warnings

If Pain Point Not Found

•Search for similar entries
•Offer to create new pain point
•Proceed with available context

Skill Metadata

Version: 1.0.0 Created: 2026-01-16 Category: Self-Improvement Integration: AI Error Learner, Skill Improver, Memory System Maintenance: On-demand (triggered by pain point escalation)