A collection of skills for Claude.ai, Claude Code, ChatGPT, and Codex, covering document processing, study material generation, PDF extraction, discussion workflows, and session continuity.
Skills are grouped by target platform. Each platform directory contains independently installable skill packages and may adapt shared workflows to the conventions of that platform.
| Skill | Purpose | Current Version | Status |
|---|---|---|---|
| extracting-pdfs | Extract and clean PDF content to markdown | - | Active |
| flashcard-generator | Generate flashcard decks from study materials | flashcard-v3.7 | Active |
| revision-notes-generator | Create concise revision notes from study materials | notes-v3.6 | Active |
| handoff | Create structured handoff documents for session continuity | - | Active |
| discuss | Discuss and clarify direction before starting work | - | Active |
| learn | Support conceptual learning and understanding | - | Active |
| Skill | Purpose | Status |
|---|---|---|
| discuss (Claude Code) | /discuss command for clarifying direction before work |
Active |
| handoff (Claude Code) | /handoff slash command for session handoff |
Active |
| resume | /resume slash command to continue from a handoff |
Active |
| Skill | Purpose | Status |
|---|---|---|
| flashcard-generator (ChatGPT) | Generate flashcard decks from study materials | Active |
| handoff (ChatGPT) | Create structured session handoffs | Active |
| learn (ChatGPT) | Support conceptual learning and understanding | Active |
| revision-notes-generator (ChatGPT) | Create concise revision notes from study materials | Active |
| Skill | Purpose | Status |
|---|---|---|
| discuss (Codex) | Discuss and clarify direction before implementation | Active |
| handoff (Codex) | Capture task state for continuation | Active |
| resume (Codex) | Resume work from a handoff document | Active |
| Skill | Purpose | Status |
|---|---|---|
| pdf-extract | PDF extraction (development version) | Archived |
| pdf-to-markdown-converter | Legacy PDF conversion | Archived |
agent-skills/
├── README.md
├── claude-ai-skills/ # Skills for Claude.ai
│ ├── flashcard-generator/
│ ├── discuss/
│ ├── extracting-pdfs/
│ ├── handoff/
│ ├── learn/
│ └── revision-notes-generator/
├── claude-code-skills/ # Claude Code slash commands
│ ├── discuss/
│ ├── handoff/
│ └── resume/
├── chatgpt-skills/ # Skills exported from ChatGPT
│ ├── flashcard-generator/
│ ├── handoff/
│ ├── learn/
│ └── revision-notes-generator/
├── codex-skills/ # Skills for Codex
│ ├── discuss/
│ ├── handoff/
│ └── resume/
└── archive/
└── claude-ai/ # Archived Claude.ai skills
├── pdf-extract/
└── pdf-to-markdown-converter/
The primary skill for extracting PDF content to clean, organized markdown. Uses a 5-step workflow (Extract, Analyse, Clean, Organise, Output) with dual extraction methods, comprehensive image handling, and reference guides for common challenges.
Full documentation: See
claude-ai-skills/extracting-pdfs/README.mdfor the complete workflow, CLI usage, dependencies, and reference guide details.
Sample prompt:
<pathname>
Use "extracting-pdfs" skill to convert this pdf into a markdown file.
Generate study flashcards from PDF or Markdown content. Card design follows evidence-based principles with six specialised card types, built-in interference checking, coverage guidance prioritised by examinability, and LaTeX equation support (KaTeX-compatible).
Current version: flashcard-v3.7
Full documentation: See
claude-ai-skills/flashcard-generator/README.mdfor design principles, card types, output format, and card design rules.
Sample prompt for claude.ai:
Create a flashcard deck from the attached study materials using the "flashcard-generator" skill.
Output the deck as a .txt file named after the source file (e.g. Physics_Chapter_5.pdf → Physics_Chapter_5.txt).
Sample prompt for Claude API:
Create a flashcard deck from the attached study materials using the "flashcard-generator" skill.
Output only the flashcard lines in the format "Question | Answer", one per line.
Do not include any preamble, headers, explanations, markdown formatting,
or code fences. The raw output will be saved directly to a text file.
Generate concise, accurate revision notes from PDF or Markdown study materials. Covers a 5-step process from reading through verification to structured markdown output, with writing guidelines prioritising conciseness, completeness, and accuracy.
Current version: notes-v3.6
Full documentation: See
claude-ai-skills/revision-notes-generator/README.mdfor writing guidelines, output format, and the full process.
Sample prompts:
Use "revision-notes-generator" skill to create revision notes of the study materials.
Create revision notes from the attached study materials using the "revision-notes-generator" skill.
Title the notes after the topic in the source, and output a .md file named after the source file (e.g. Physics_Chapter_5.pdf → Physics_Chapter_5.md).
Create structured handoff documents for seamless continuity across Claude conversations. When a session ends, the handoff captures the full working state — objective, decisions with rationale, failed approaches, and next steps — so a fresh conversation can continue without re-explanation.
The skill adapts its structure and depth to the type of work (research, writing, planning, building, learning, problem-solving) and the complexity of the session.
Full documentation: See
claude-ai-skills/handoff/README.mdfor work types, depth tiers, document structure, and usage instructions.
Pause before starting work to clarify requirements, surface assumptions, and agree on direction.
Support requests for intellectual understanding with explanations adapted to the learner and topic.
Slash commands designed for the Claude Code CLI. These are installed at user-level (~/.claude/skills/) and invoked with /command syntax.
A /discuss command that pauses implementation so requirements and direction can be clarified first.
A /handoff slash command for Claude Code that generates structured handoff documents optimised for coding sessions. Captures implementation decisions, failed approaches, error messages, and file-level state that the codebase alone does not preserve. Each handoff references the previous one, creating a traceable handover chain across sessions.
Full documentation: See
claude-code-skills/handoff/README.mdfor work types, depth tiers, the handover chain, and installation instructions.
A /resume slash command that loads a handoff document and gets you productive immediately. Companion to /handoff — it finds the most recent handoff (or a specific one you name), reads it, reviews the key files listed in it, and continues from the next step.
Full documentation: See
claude-code-skills/resume/README.mdfor usage and the handoff/resume workflow.
The chatgpt-skills/ directory contains the skill packages uploaded to and exported from ChatGPT, including the generated agents/openai.yaml metadata and icons.
The collection currently includes Flashcard Generator, Handoff, Learn, and Revision Notes Generator.
The codex-skills/ directory contains Codex adaptations of the coding workflow skills. They are installed under ~/.codex/skills/ and can be invoked by name, such as $handoff and $resume.
The collection currently includes Discuss, Handoff, and Resume.
The following Claude.ai skills have been moved to archive/claude-ai/. They are preserved for reference but are no longer actively maintained.
Location:
archive/claude-ai/pdf-extract/
The development/original version of the PDF extraction skill. Contains the core Python extraction script.
Note: For production use, see extracting-pdfs which is the current active version.
Features:
- Same extraction capabilities as extracting-pdfs
- Contains
extract_pdf.pyscript (1,500+ lines) - Full reference documentation included
Location:
archive/claude-ai/pdf-to-markdown-converter/
The original PDF conversion skill using visual PDF understanding.
Status: Deprecated. Superseded by the more sophisticated
extracting-pdfsskill.
- Extract everything — No hardcoded cleanup rules during extraction
- Preserve raw content — Keep data intact for intelligent post-processing
- Rich metadata — Provide comprehensive context for document understanding
- Manual over automated — Complex decisions handled manually for better results
- Atomic flashcards — One fact per card for effective learning
- Accuracy first — Verify and correct information in study materials
- Structured continuity — Capture working state, not conversation summaries, so sessions can resume without re-explanation
- Python 3 — Core scripting language
- PyMuPDF (fitz) — Low-level PDF reading and image extraction
- PyMuPDF4LLM — Enhanced markdown formatting with table support
- YAML/JSON — Metadata formats
- Markdown — Output format
This repository uses per-skill version labels. The current study-generation labels are flashcard-v3.7 for flashcard-generator and notes-v3.6 for revision-notes-generator. See Releases for version history.
This work is licensed under CC BY 4.0 - you're free to share and adapt with attribution.