Patternify

Transform productive workflows, tool orchestrations, and decision heuristics into reusable Claude Code components.

Instructions

• Consider the recent conversation history and any patterns that emerged.
• Specific user instructions should be followed unless they are contradictory to the task at hand. $ARGUMENTS

Steps

1. Load — Use the Skill tool and load the baselayer:patternify skill
2. Consider — Ultrathink and analyze the conversation, identify candidate patterns, and assess reusability
3. Dispatch or Execute — Choose execution path based on available tools:
   • If Task tool available: Dispatch the baselayer:pattern-analyzer subagent in background mode, passing along the pattern hint (if any) and conversation context summary
   • If Task tool unavailable: Execute the pattern analysis methodology directly using the loaded skill
4. Monitor — If subagent dispatched, use TaskOutput to retrieve structured JSON findings
5. Generate — Based on findings, create the appropriate component files

Context Handoff (for subagent dispatch)

When dispatching to the pattern-analyzer subagent, include:

• Pattern hint from arguments (if provided)
• Summary of productive workflows from conversation
• Key tool sequences and decision points observed
• User satisfaction signals ("that worked great")

Pattern Types

• Workflow: Multi-step process with phases
• Orchestration: Tool coordination toward a goal
• Heuristic: Decision rule with exceptions

Component Mapping

Is it a multi-step process?
├─ Yes → Requires judgment? → SKILL (not fully automatable)
│        Fully automatable? → COMMAND (scripted)
└─ No → Event-triggered? → HOOK
         Requires expertise? → AGENT
         User-initiated? → COMMAND

Output Locations

• Plugin: **/{plugin}/skills/{name}/ — shareable via marketplace
• Project: {CLAUDE_PROJECT_DIR}/.claude/skills/{name}/ — team-shared
• Personal: ~/.claude/skills/{name}/ — global personal toolkit

Quality Criteria

Before generating, validate:

• Specific: Clear trigger and scope
• Repeatable: Works across contexts
• Valuable: Worth the overhead (saves >5 min)
• Documented: Others can understand
• Scoped: Single responsibility

Skip if: <3 occurrences, context-dependent, simpler inline.