skill-factory

name	skill-factory
version	1.0.0
category	coding
description	Automatically extract reusable skills from complex workflows. Use when: (1) you just completed a multi-step task that could recur, (2) you notice a repeated pattern across sessions, (3) the user says 'make this a skill' or 'remember how to do this'. NOT for: one-off tasks, simple operations, or tasks already covered by existing skills.
metadata	{"deneb":{"emoji":"🏭","tags":["automation","extraction","pattern","procedural-memory"],"related_skills":["evolution-proposal","skill-creator","skill-evolution"]}}

Skill Factory — Autonomous Skill Extraction

Inspired by hermes-agent's procedural memory: the agent creates skills from experience.

When to Use

Trigger autonomously after completing a complex workflow (5+ tool calls, multi-step reasoning):

1. You just solved a non-trivial problem with a reusable pattern
2. You notice you've done a similar workflow before (check session-logs)
3. The user explicitly asks to capture a workflow as a skill

Do NOT create a skill when:

• The task is a one-off with no reuse potential
• An existing skill already covers the workflow
• The pattern is too specific to generalize

Extraction Procedure

Step 1: Identify the Pattern

Review the just-completed workflow. Ask:

• What was the goal?
• What tools/commands were used?
• What decisions were non-obvious?
• What pitfalls did I encounter?
• Would a future agent benefit from these instructions?

Step 2: Choose Category

If the skill is about...	Category
Code, builds, testing, git, CI	coding
Daily workflows, docs, summarization	productivity
Server, monitoring, infra	devops
External APIs, bridges, imports	integration

Step 3: Generate SKILL.md

Create skills/<category>/<name>/SKILL.md with:

---
name: extracted-skill-name
version: "1.0.0"
category: <category>
description: "<What it does>. Use when: <triggers>. NOT for: <anti-patterns>."
metadata:
  {
    "deneb":
      {
        "emoji": "<icon>",
        "tags": ["keyword1", "keyword2"],
        "related_skills": ["existing-skill"],
      },
  }
---

Step 4: Write Body Sections

Follow the standard body structure:

1. When to Use — Concrete trigger conditions from the workflow
2. Quick Reference — The key commands/patterns extracted
3. Procedure — Step-by-step from the actual workflow, generalized
4. Pitfalls — Errors encountered and how to avoid them
5. Verification — How to confirm the workflow succeeded

Step 5: Validate

• Does the skill add value beyond what's already in existing skills?
• Is the description specific enough for the LLM to know when to load it?
• Are the tags useful for discovery?
• Is the procedure generalizable (no hardcoded paths/values)?

Nudging Protocol

After completing a complex task, briefly consider:

"이 작업은 스킬로 만들 가치가 있는가?"

If yes, offer to the user: "이 워크플로우를 스킬로 저장할까요?" If the user agrees, execute the extraction procedure.

Anti-Patterns

• Do NOT create skills that duplicate existing tool functionality
• Do NOT create skills with hardcoded credentials, paths, or user-specific data
• Do NOT create overly broad skills ("do everything") — narrow scope, deep quality
• Do NOT create skills for tasks the LLM already handles well without instructions