Pattern Detection & Automation Suggestion Skill

When to Trigger (Auto-Detect)

Monitor for these patterns during sessions:

Command Repetition

• Same bash command run 3+ times
• Same sequence of commands repeated
• Similar commands with slight variations

Task Repetition

• Same type of request across sessions
• Repeated file modifications in same pattern
• Recurring debugging steps

Knowledge Gaps

• Information I needed but wasn't in CLAUDE.md
• Corrections from user about project conventions
• Confusion that required clarification

Workflow Patterns

• Multi-step processes done manually
• Copy-paste patterns
• Repeated context explanations

Detection Thresholds

Pattern Type	Threshold	Action
Same command	3x in session	Suggest command
Similar commands	5x in session	Suggest parameterized command
Same task type	2x across sessions	Suggest skill
User correction	1x	Suggest adding to CLAUDE.md
Missing info	2x	Suggest documenting

Suggestion Format

When pattern detected, present:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
PATTERN DETECTED

I noticed: [specific observation]
Frequency: [how often]

Suggested automation:
[brief description of what would be created]

Options:
1. Create project command (.claude/commands/)
2. Create global command (~/.claude/commands/)
3. Create project skill (.claude/skills/)
4. Add to project CLAUDE.md
5. Add to global ~/.claude/CLAUDE.md
6. Skip for now
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Automation Templates

For Repeated Commands -> Create Command

.claude/commands/[name].md

:

# [Command Name]

[Description of what this does]

## Steps
1. [step]
2. [step]

## Usage
/project:[name] [optional args]

For Complex Workflows -> Create Skill

.claude/skills/[name]/SKILL.md

:

---
name: [name]
description: [what it does]
---

# [Skill Name]

## When to Use
[trigger conditions]

## Steps
1. [step]
2. [step]

## Output
[expected result]

For Conventions/Rules -> Add to CLAUDE.md

## [Section Name]
- [rule or convention]
- [example]

For Personal Preferences -> Add to CLAUDE.local.md

## My Preferences
- [preference]

Post-Creation

After creating automation:

1. Confirm with user
2. Test if applicable
3. Summarize what was created and how to use it

Created: [file path]
Invoke with: [command or trigger]

Examples

Example 1: Repeated Test Command

Pattern: `npm test -- --watch src/utils` run 4 times

Suggestion: Create command /project:test-utils
Content: Runs tests in watch mode for utils directory

Example 2: Repeated Debugging Steps

Pattern: Same sequence: check logs -> read config -> verify env

Suggestion: Create skill /debug-config
Content: Systematic config debugging workflow

Example 3: User Correction

Pattern: User said "always use async/await, never .then()"

Suggestion: Add to CLAUDE.md code conventions
Content: "Use async/await for promises, avoid .then() chains"

Self-Improvement Loop

At end of significant tasks, briefly evaluate:

1. What information would have helped?
2. What steps were repetitive?
3. What could be automated next time?

If improvements identified, suggest them proactively.