Skip to content
Skills Library
AI & agents · Analysis

skill-extractor

/skill-extractor

Extract actionable Claude Code skills from raw source material — transcripts, conversations, workflows, expertise dumps. This skill identifies repeatable, promptable workflows embedded in content and scores them by leverage. Use when processing a corpus (podcast transcripts, blog posts, course material) to discover what skills could be built from it.

Repository
skill-stack ↗
Licence
MIT
Last repository change
07.07.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 07.07.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Extract actionable Claude Code skills from raw source material — transcripts, conversations, workflows, expertise dumps. This skill identifies repeatable, promptable workflows embedded in content and scores them by leverage. Use when processing a corpus (podcast transcripts, blog posts, course material) to discover what skills could be built from it.

Original author description

Extract actionable Claude Code skills from raw source material — transcripts, conversations, workflows, expertise dumps. This skill identifies repeatable, promptable workflows embedded in content and scores them by leverage. Use when processing a corpus (podcast transcripts, blog posts, course material) to discover what skills could be built from it.

How it works

  1. The Core Distinction
  2. Skill Taxonomy
  3. Available Tool Ecosystem
  4. Extraction Workflow
  5. Routing Non-Skills
  6. Test Corpora
  7. Related Skills
Original SKILL.md headings · 174 lines, 1,028 words
SKILL.md preview
---
name: skill-extractor
description: Extract actionable Claude Code skills from raw source material — transcripts, conversations, workflows, expertise dumps. This skill identifies repeatable, promptable workflows embedded in content and scores them by leverage. Use when processing a corpus (podcast transcripts, blog posts, course material) to discover what skills could be built from it.
---

# Skill Extractor

Turn raw source material into a list of buildable Claude Code skills. Not a knowledge base — skills are *workflows Claude can execute*, not facts Claude should know.

## The Core Distinction

**Skill** = a repeatable workflow with inputs, steps, and outputs that Claude executes inside a coding environment. Writing a YouTube script. Running an SEO audit. Drafting a newsletter. Creating ad copy.

**Not a skill** = static knowledge, reference material, domain expertise. How to raise meat birds. The history of fasting. Nutritional science. These become wiki articles, books, or reference docs — not skills.

**The test:** Can Claude *do* this thing, repeatedly, with different inputs, and produce a useful output? If yes → skill. If it's something Claude *knows* and references → not a skill.

---

## Skill Taxonomy

Score each candidate against these six hallmarks. A strong skill exhibits 2+ of these.

### 1. Repeated SOP / Workflow
Something done regularly as part of a content or business process. The more often it's done, the higher the leverage.

**Examples:** Newsletter drafting, social post creation, podcast show notes, SEO content briefs, weekly reporting

### 2. Disposable One-Shot
Valuable but infrequent. Done once per project or client. Still worth codifying because it saves hours when needed — and becomes an agency offering if done for others.

**Examples:** Writing a LinkedIn bio, Amazon category research, brand identity creation, book launch checklist, landing page copy

### 3. Specific Knowledge Applied as Process
Domain expertise distilled into a repeatable method — not raw knowledge, but knowledge *operationalized* into steps Claude can follow.

**Examples:** Kallaway's "7 Lego Bricks" for short-form → YouTube scriptwriting skill. Dan Koe's "one-person business" model → content strategy skill. NOT "the history of short-form video" (that's reference).

### 4. Step-by-Step Process
A clear sequence where order matters. Often teachable, often already described as numbered steps in the source material.

**Examples:** Video editing workflow, podcast production pipeline, email sequence writing, ad creative testing process

### 5. Example-Driven
Skills where having 3-5 concrete examples dramatically improves output quality. The examples ARE the skill — they teach by pattern, not by instruction.

**Examples:** Voice/style skills (Trung Phan, Tyler Cowen), ad creative frameworks, hook writing, cold open creation

### 6. Tool-Augmented
Skills that benefit from MCP servers, API calls, CLI tools, or other integrations. These extend Claude's autonomy — the model can work longer without human input.

**Examples:** SEO keyword research (DataForSEO), social media scraping (Apify), RSS curation, image generation (Gemini), video processing (ffmpeg)

---

## Available Tool Ecosystem

When evaluating whether a skill candidate is tool-augmented, consider what's currently available:

**MCP Servers (live in this workspace):**
- Slack (channels, messages, search)
- Google Drive (docs, sheets, slides, folders)
- Google Calendar (events, scheduling, freebusy)
- Apify (web scraping actors — YouTube, Twitter, any site)
- Notion (pages, databases, search)
- Video-audio (trim, convert, subtitles, overlays, concatenate)
- Context7 (library documentation lookup)

**CLI Tools:**
- ffmpeg (video/audio processing)
- yt-dlp (YouTube downloading)
- Whisper (transcription)
- qmd (local markdown search / RAG)
- gh (GitHub CLI)
- Bun/Node (JS execution)
- Python (scripting, data processing)

**APIs (via scripts):**
- DataForSEO (keyword research, SERP, rankings)
- Gemini (deep research, image generation, large-context writing)
- ElevenLabs (voice cloning)
- HubSpot (email marketing)
- Webflow (CMS publishing)

When a source mentions a workflow that *could* be automated with these tools but the creator does it manually, that's a high-value skill candidate.

---

## Extraction Workflow

### Phase 1: Ingest and Scan

**Input:** Raw source material — transcripts, blog posts, course outlines, conversation logs, wiki chunks.

Read the material. Identify skill candidates based on pattern recognition from existing skills in this workspace. For each candidate:

```markdown
### [Candidate Name]
- **What it does:** One sentence
- **Source:** Where in the material this was found
- **Hallmarks:** Which taxonomy items (1-6) it exhibits
- **Input → Output:** What goes in, what comes out
- **Frequency:** daily / weekly / per-project / one-time
```

### Phase 2: Score and Rank

| Dimension | Question | Score |
|-----------|----------|-------|
| **Leverage** | How much time/effort does this save per use? | 1-5 |
| **Frequency** | How often would this be used? | 1-5 |
| **Promptability** | How well can Claude execute this with a skill file? | 1-5 |

**Leverage × Frequency × Promptability = Extraction Priority**

High-priority (50+): build immediately. Medium (25-49): build when needed. Low (<25): note or discard.

### Phase 3: Spec the Winners

For each high-priority candidate, draft a skill spec for the skill-creator:

```markdown
## Skill Spec: [name]

**Purpose:** What this skill accomplishes
**Trigger:** What would the user say to invoke this?
**Not for:** What this skill should NOT be used for

**Input → Output**
**Workflow:** [numbered steps]
**Bundled Resources:** scripts, references, assets needed
**Tool Dependencies:** MCPs, APIs, CLI tools required
**Examples Needed:** What examples would make this work well?
**Source Attribution:** Where this was extracted from
```

### Phase 4: Present to User

1. Top 5 high-priority skills with full specs
2. Medium-priority as one-liners
3. Non-skill material routed elsewhere

Ask: "Which should I build first?"

---

## Routing Non-Skills

| Type | Destination |
|------|-------------|
| Domain knowledge / facts | Wiki article or reference doc |
| Opinion / philosophy | Blog post or book chapter |
| Personal story / anecdote | Narrative snippet for content |
| Tool recommendation | Relevant project's CLAUDE.md |
| Business strategy | Strategy doc or CIA-OFFER.md |

---

## Test Corpora

Already chunked and indexed in wiki-projects:
- **Kallaway** — 66 chunks, framework-dense → should yield many skills
- **Jenny Hoyos** — 60 chunks, short-form methodology → should yield skills
- **Colin & Samir** — 119 chunks, strategy + interviews → mixed

---

## Related Skills

- **skill-creator** — Takes a spec and builds the full skill. Extractor feeds into creator.
- **voice-analyzer** / **voice-wizard** — Specialized extractors for voice/style skills.
- **anti-ai-writing** — Example of a well-built skill.
- **book-chapter-writer** — Example of a complex multi-phase workflow skill.

---

*Extract the workflow, not the knowledge. If Claude can DO it repeatedly with different inputs, it's a skill. If Claude can only KNOW it, it's a reference doc.*

Versions

RepositoryScoreLicence
This page · cdeistopened/skill-stack100MIT
trailofbits/skills-curated100CC-BY-SA-4.0

Database snapshot: 13.11.2026 · package version 305c23d9a303 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100conductorscore
    Score your Claude Code skill — measure agent leverage, craft, and anti-patterns from your local transcripts. Use when the user types /conductorscore.
  • 90search-skill
    Discover and evaluate Claude Code skills and plugins from curated sources. Use when searching for existing skills/plugins, finding inspiration, or before creating a new skill. Triggers include "search skill", "find skill", "find plugin", "skill for X", "what skills exist for", "discover skills".
  • 100skill-lint
    Builds, audits and scores portable Agent Skills (SKILL.md plus scripts) for Claude Code, Codex and Cursor. Use when the user asks to write a skill, convert a workflow into a reusable skill, fix a skill that never triggers, or review a SKILL.md before publishing.
  • 100skillproof
    Lints and scores Claude Code Agent Skills with SkillProof. Use when the user asks to validate a skill, diagnose SKILL.md warnings, or add skill checks to CI.
  • 95writing-claude-skills
    Use when user asks to create, write, edit, or test a skill. Also use when documenting reusable techniques, patterns, or workflows for future Claude instances.
  • 100source-to-skill
    Use when compiling raw knowledge sources (transcripts, docs, PDFs, books, video notes) into structured, operational agent skills.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.