Stats
Links
Categories
Strict TDD skill for Claude Code
npx claudepluginhub mdodkins/claude-tdd-skillStrict Red-Green-Refactor TDD workflow for all code-writing tasks
A custom skill that puts Claude into strict Test Driven Development mode, enforcing the Red-Green-Refactor cycle for every behaviour implemented.
You must be authenticated with the GitHub CLI and have git configured to use it for HTTPS credentials:
gh auth login
gh auth setup-git
If you're already logged in but using SSH, switch to HTTPS and configure the credential helper:
gh auth login --hostname github.com --git-protocol https
gh auth setup-git
Register this repo as a plugin marketplace, then install:
/plugin marketplace add mdodkins/claude-tdd-skill
/plugin install tdd@mdodkins-tdd
Copy the skill file into your personal skills folder:
mkdir -p ~/.claude/skills/tdd
cp skills/tdd/SKILL.md ~/.claude/skills/tdd/SKILL.md
Or for a specific project, copy it into the project's .claude directory and commit it:
mkdir -p .claude/skills/tdd
cp /path/to/tdd-skill/skills/tdd/SKILL.md .claude/skills/tdd/SKILL.md
In any Claude Code session, ask:
What skills are available?
You should see tdd in the list. You can also invoke it directly with /tdd.
When triggered, this skill instructs Claude to:
The skill auto-triggers when you ask Claude to implement, build, create, fix, refactor, or write code.
It does not trigger for debugging, questions about code, exploratory/spike work, or writing tests for already-existing code.
This skill was developed with reference to Anthropic's official skill authoring guidance. Key decisions:
The official best practices state: "Write using verb-first instructions, not second person."
This isn't just style — the skill-creator skill notes that imperative form is clearer for AI consumption and maintains consistency.
The skill-creator skill advises: "Try hard to explain the why behind everything... If you find yourself writing ALWAYS or NEVER in all caps, that's a yellow flag — reframe and explain the reasoning so that the model understands why the thing you're asking for is important."
Claude has good theory of mind. Understanding why a rule exists leads to better adherence than being commanded to follow it.
The official docs emphasise: "Claude is already very smart. Only add context Claude doesn't already have." Explanations of well-known concepts (like Arrange-Act-Assert) were removed because they consume tokens without adding value.
Skill descriptions are the primary triggering mechanism and share a combined character budget (~2% of context window, ~16K chars fallback) across all installed skills. A bloated description is more costly than a longer body, because descriptions are always loaded while the body only loads on trigger.
The description was written to be:
The skill body is ~150 lines / ~1,000 words. The recommended limit is <500 lines. Progressive disclosure (splitting into reference files) is unnecessary at this size. If the skill grows significantly, detailed content should move to a references/ directory.
| Metric | Limit | This Skill |
|---|---|---|
| SKILL.md body | <500 lines | ~150 lines |
| Description field | Max 1024 chars | ~500 chars |
| All skill descriptions combined | ~2% of context window (~16K chars) | N/A |
The body is only loaded into context when the skill triggers. The description is always in context. This means description brevity matters more than body brevity for overall context efficiency.
Development marketplace for Superpowers core skills library
Harness-native ECC skills, hooks, rules, MCP conventions, and operator workflows
Open Design — local-first design app exposed to coding agents over MCP. Install once with your agent's plugin command and projects/files/skills are reachable through stdio.