skills

A growing collection of Claude Code skills, plus a small plugin for persistent persona modes.

Contents

Name	Type	What it does
debate-team	Skill	Spin up distinct AI personas to debate a topic, then synthesize a verdict.
design-md	Skill	Build a DESIGN.md design-token spec from a brief or an existing codebase.
modes/	Plugin	Hook-driven persistent personas (ships contrarian).

• Standalone skills
• modes/ — persistent persona plugin
• Adding your own
• Attribution

Standalone skills

Each is a directory with a SKILL.md (YAML frontmatter + instructions) and optional references/. Claude loads one on demand when its description matches the task. Self-contained — no hooks or scripts.

debate-team

Spins up a roster of distinct AI personas that argue a topic via inter-agent messaging, then delivers a moderated synthesis. Auto-casts the panel from the topic — customers debating a product, engineers debating a stack, stakeholders arguing strategy, multi-angle devil's advocate.

• Use for: debate, stress-test, pressure-test, red-team, or roundtable on a decision, plan, or product.
• Ships with: references/casting-library.md — persona archetypes the roster is cast from.
• Requires: a Claude environment with multi-agent support (TeamCreate / SendMessage).

design-md

Builds a DESIGN.md file — the design.md spec: YAML design tokens plus human-readable rationale that gives agents a persistent, structured understanding of a design system.

• Brainstorm — walk a greenfield project through brand / color / type / layout choices to full coverage.
• Consolidate — extract real tokens from an existing codebase's CSS / Tailwind / theme files.
• Ships with: references/spec.md (token schema + section rules) and references/example.md (worked reference).

Install

# user-level (available in all projects)
cp -R debate-team ~/.claude/skills/

# or project-level
cp -R debate-team /path/to/project/.claude/skills/

Then invoke with /<skill-name>, or let Claude trigger it automatically from the skill's description.

modes/ — persistent persona plugin

A small, hook-driven plugin for personas that stay active on every turn until you switch them off — something a plain skill can't do, since a skill only applies to the turn it's invoked on.

Ships one mode today:

• contrarian — a sharp, skeptical advisor that pressure-tests every decision, plan, or claim instead of validating it. Names the load-bearing assumption, fires three concrete counterarguments, proposes a superior alternative, surfaces the blind-spot risk, and ends with a verdict: Proceed / Reconsider / Stop.

Why it needs hooks (not just a skill)

/contrarian works as a one-shot skill on its own, but the persistence — re-asserting the persona on every prompt and honoring stop contrarian immediately — is driven by two hooks:

• hooks/modes-tracker.js (UserPromptSubmit) — detects on/off triggers, manages the ~/.claude/.modes-active flag, and re-injects the persona each turn.
• hooks/modes-activate.js (SessionStart) — injects the persona up front on session start / resume / clear / compact when a mode is already active.

Hooks swallow all errors and exit 0 — they can never block a prompt or a session start.

Install

cd modes
./install.sh      # copies hooks + skill to ~/.claude, registers the hooks in settings.json (idempotent, backed up)
./test.sh         # optional self-test in a sandbox HOME
./uninstall.sh    # clean removal

Requires Node ≥ 18. Then:

Type this	Effect
/contrarian or contrarian mode	activate
/contrarian off or stop contrarian	deactivate contrarian
normal mode	deactivate whatever mode is active

See modes/README.md for the full mechanism and how to author new modes.

Adding your own

• A standalone skill: create <name>/SKILL.md with name + description frontmatter, then cp -R it into a skills directory.
• A persistent mode: create modes/skills/<name>/SKILL.md with mode: true in its frontmatter and re-run modes/install.sh. Details in modes/README.md.

Attribution

design-md implements the open google-labs-code/design.md spec (alpha).