CookiesForAI

writing-skills

Writing Skills

Writing Skills is an agent skill from Superpowers: your agent writes new skills test-first, running pressure scenarios with subagents before and after the skill exists.

Agent skill by Jesse Vincent · Updated

Install Writing Skills

In Claude Code, install the whole Superpowers plugin from Anthropic's official marketplace

Terminal
/plugin install superpowers@claude-plugins-official

Or install only this skill with the skills CLI

Terminal
npx skills add obra/superpowers --skill writing-skills

Or paste this into your coding agent: Install the agent skill writing-skills from github.com/obra/superpowers. A skill can include scripts that run on your computer, so read its source first.

What Writing Skills does

Writing Skills is one of the skills in Superpowers, Jesse Vincent's open-source method for AI coding agents. It loads when creating a skill, editing one, or checking that a skill works before deployment. It treats skill writing as test-driven development applied to documentation, and requires an understanding of the Test-Driven Development skill first. Its rule mirrors that one: no skill, and no edit to a skill, without a failing test first.

The test in Writing Skills is a pressure scenario run with a subagent. The agent first runs it without the skill and records what the subagent does and the excuses it gives, then writes the smallest skill that addresses those failures, runs the scenario again to confirm the subagent complies, and closes new loopholes. For wording, it runs micro-tests: five or more samples per variant against a control with no guidance, reading every flagged match by hand.

Writing Skills also covers structure: a name of letters, numbers and hyphens, and a description that starts with "Use when" and names only the triggering conditions, since agents may follow a description that summarizes the workflow instead of reading the skill. It sets word targets for skills that load often. Bundled files include Anthropic's skill authoring best practices, a guide to testing skills with subagents, and a script that renders a skill's flowcharts to SVG.

When to use Writing Skills

  • You are writing your own skill and want proof that it changes how the agent behaves.
  • Your skill exists but the agent skips it or finds a way around its rules.
  • You want a skill description that makes the agent load the skill at the right time.
  • You are editing an existing skill and want to check it still works.

When to pick something else

  • You want graded benchmarks, a review page and description tuning: use Skill Creator from Anthropic.

Which agents Writing Skills works in

Jesse Vincent documents Writing Skills for Claude Code, Codex (app and CLI), Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, Antigravity, Devin CLI, Factory Droid, Grok Build CLI, Kimi Code, Pi, Qwen Code, Hermes Agent, Muse. The open skills CLI also installs it into 78 agents, including Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode (we listed it with the CLI on October 1, 2026). See where each agent looks for skills.

Superpowers is written in Claude Code's vocabulary (subagent dispatch, todos, the Skill tool) and ships tool mappings for Codex, Gemini CLI, Pi, Antigravity, Hermes Agent and Muse. Installed as the full plugin, a session-start hook loads Using Superpowers so the agent reaches for skills on its own; one skill installed with the skills CLI gets no hook and may point to sibling skills you do not have. This skill needs a subagent tool: Claude Code's Agent tool, or in Codex the multi-agent feature switched on in config.toml (multi_agent = true).

Writing Skills license

Writing Skills is published under MIT.

  • MCP Builder: MCP Builder is Anthropic's open-source agent skill for building MCP servers in TypeScript or Python, from API research and tool design to testing and a 10-question evaluation. (179,326 repository stars on October 2, 2026)
  • Skill Creator: Skill Creator is Anthropic's open-source agent skill for making skills: the agent drafts a SKILL.md with you, tests it with and without the skill, and tunes its description so it triggers. (179,326 repository stars on October 2, 2026)

More from Jesse Vincent

  • Brainstorming: Brainstorming is an agent skill from Superpowers that makes your coding agent work out what you want and agree a design with you before it writes any code. (293,976 repository stars on October 2, 2026)
  • Diagnosing Superpowers: Diagnosing Superpowers is an agent skill from Superpowers that reads a session's transcripts when work went wrong and reports what happened, with every finding tied to a file and line. (293,976 repository stars on October 2, 2026)
  • Dispatching Parallel Agents: Dispatching Parallel Agents is an agent skill from Superpowers: when several problems are independent, your agent sends a focused subagent to each at once, then checks the results together. (293,976 repository stars on October 2, 2026)
  • Executing Plans: Executing Plans is an agent skill from Superpowers in which your agent works through a written plan itself, task by task with tests, then gets one fresh review of the whole branch. (293,976 repository stars on October 2, 2026)

Lists that include writing-skills