Claude Skills

The Zoom Skill provides an interface to manage Zoom resources via the Zoom API, supporting both JSON and Markdown output formats. Key capabilities include:

• Meeting Management: List, create, update, and delete Zoom meetings (both instant and scheduled).
• Recording Access: Retrieve cloud recordings, including their transcripts, summaries, and download links.

Claude Skills is a collection of ~100 skills designed for Claude Code. These skills enable advanced capabilities such as meeting pipelines, research, image generation, TDD, publishing, and personal analytics.

Each skill is structured as a self-contained folder containing a SKILL.md file. Claude triggers these skills based on natural language instructions provided by the user.

The confide:annotate skill is used to build and verify a Personally Identifiable Information (PII) gold set using human annotators. It provides a workflow for labeling PII spans in transcripts, measuring inter-annotator agreement (IAA) using Cohen's or Fleiss' kappa (κ), and adjudicating labels to create a final gold standard.

Key Privacy Invariants:

• Use synthetic or consented data only. If using real data, anonymize it first using confide:anon.
• Names and transcript text stay local. Annotators work on their own machines; only exported label files and statistical results (κ, F1, disagreement clusters) are shared with the coordinator.
• No PII is re-shared during the scoring or adjudication process.

The decision-toolkit is a skill designed to generate structured decision-making tools (interactive guides, bias checkers, scenario explorers, and dashboards) to help humans navigate significant choices. It follows a philosophy of guiding rather than deciding, using progressive disclosure to reduce cognitive load.

When to use

Invoke this skill for:

• Collaboration/partnership decisions
• Career or job changes
• Investment of significant time/money
• Project prioritization
• Technology/tool selection
• Any choice with multiple factors and uncertainty

Do not use for: Trivial decisions, emergency responses, or when you only need information.

The design-tokens skill is a deterministic core built as a dependency-free Python package named dtokens. It is designed to validate, merge, resolve, and export Design Token Community Group (DTCG) token files into CSS.

Core Functionality:

• Scaffolding/Validation: Uses setup-edit to scaffold and validate token files.
• Resolution: Uses use to resolve tokens to CSS and an LLM-readable context file.
• CLI: Provides a tokens CLI (via a tokens shim) for interacting with the package.

Architecture: The dtokens package is organized into modules by responsibility: model, validate, merge, resolve, export_css, and cli.

The llm-cli skill allows you to process text, files, and multimedia using various LLM providers (OpenAI, Anthropic, Google Gemini, and Ollama) via the llm CLI tool. It supports both one-off non-interactive commands and interactive chat sessions.

Supported Providers & Aliases:

• OpenAI (openai, gpt): e.g., gpt-4o, o3-mini
• Anthropic (anthropic, claude): e.g., claude-sonnet-4.5, claude-3.5-haiku
• Google Gemini (google, gemini): e.g., gemini-2.5-pro, gemini-2.5-flash
• Ollama (ollama, local): e.g., llama3.1, mistral-large-2 (runs locally)

# Non-interactive: Process text
llm "Your prompt here"

# Non-interactive: Use a specific model
llm --model gpt-4o "Process this text"

# Non-interactive: Pipe content
cat document.md | llm "Summarize"

# Interactive mode
llm --interactive

The pre-session-portrait skill helps consultants build a compressed, visualizable brief of a client before a session. It uses a 7-lens JTBD-inspired interview to turn open-ended client answers into a structured 11-field YAML portrait.

When to use it:

• Preparing for an upcoming client call.
• When a user requests an "intake", "portrait interview", or "questions before our session".
• When you need a reusable client-intake instrument.

Design Principle: "Rich in, compressed out." The client provides free-form answers (voice-note friendly), and the skill compresses them into a single-line-per-lens format.

The tufte-report skill creates standalone, publication-quality HTML reports inspired by Edward Tufte's design principles. It is designed for high information density, minimal 'chart junk', and a typography-first approach.

Use this skill when you want to create:

• Data reports or activity dashboards
• Infographics
• Personal analytics pages or health tracker visualizations
• Any document combining narrative text with interactive charts and tables.

Trigger phrases include:

• "make a report like Tufte"
• "create an infographic"
• "build a dashboard"
• "visualize my data"