Skill Creator
Scaffold a new AI agent skill from a plain-language job description: folder, SKILL.md frontmatter, trigger description, and a first draft of the instructions.
Get the complete skill folder, including its scripts and reference files.
·
When to use this skill
Use Skill Creator when you know the job an agent should repeat (release notes, PDF fill, repo triage) but you do not yet have a SKILL.md. It is the right starting point for a new Claude, Cursor, or Codex skill rather than pasting a one-off prompt into chat.
When not to use it
Skip Skill Creator when the job is a one-off request with no reusable workflow to capture. It supports both creating and improving skills; use MCP Builder when you need to implement a callable service rather than write agent instructions.
How to install
Where will you use it?
Install the publisher’s plugin
This installs a bundle that can include other skills. Choose this method or the manual alternative below; you only need one installation.
Run these commands in Claude Code:
Claude Code CLI/plugin marketplace add anthropics/skills /plugin install example-skills@anthropic-agent-skills
- Ask Claude Code to use Skill Creator on a small task and confirm that it loads. Check the setup requirements in the publisher’s guide if it cannot run.
Alternative: install the folder manually
Use these steps if you choose the manual folder method. If you installed the plugin above, skip this alternative.
- Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/skill-creator/SKILL.md. For a shared project, commit the folder at .claude/skills/skill-creator/ instead.
- Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
- For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.
What supports this recommendation?
76/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.
- Documentation
- 77/100
- Practical detail
- 68/100
- Source evidence
- 70/100
- Source signals
- 96/100
Model compatibility evidence
- claudeinferred
The skill text mentions Claude or a closely associated term. This is keyword-derived compatibility; no hands-on model test is recorded.
- chatgptinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- geminiinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- copilotinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- llamainferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- perplexityinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- mistralinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
- grokinferred
The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.
Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.
Try one small task
Best first exercise: create a narrow skill and compare its output on a few explicit cases.
See actual outputs and execution limits →What you need
- An agent with file-writing access and the complete skill-creator directory, including its evaluation tools.
- A recurring task, representative input, and criteria that distinguish a good output from a plausible one.
Draft a meeting-action skill with positive and negative trigger cases.
Use skill-creator to make a meeting-actions skill. It should extract owner, task and due date from meeting notes, mark missing fields unknown, and never invent actions. Create SKILL.md plus three test inputs: two meeting records and one unrelated recipe. Define expected behavior for each. Show the draft and how to compare runs with and without the skill; label tests unrun until you actually execute them.
Expected result
A specification to check against, not output from an AI run.
A skill folder and three evaluation cases that include an unrelated request and a missing-owner example.
- SKILL.md has a name and a description that identifies the intended task.
- The missing owner stays unknown and the unrelated recipe does not produce meeting actions.
- Any claim of improvement includes actual run artifacts; unexecuted cases remain labeled as plans.
Where this exercise stops
- A well-formed SKILL.md does not establish that an agent will select it reliably.
- Upstream supports creation, editing, and evaluation; quality still depends on representative tests.
Recorded example run
These files came from actual execution in a Codex session. They were produced and checked within the same session; this is scoped evidence, not independent review or a measured success rate.
Meeting-action skill: author-applied examples
Executed · Automated Codex exercise
Codex agent; Node.js 24.13.0, Python 3.12.3, Linux 7.0.0-1010-aws. Exact model version not exposed.
Read pinned skill-creator instructions, authored a narrow SKILL.md and three evaluation inputs, then explicitly applied the draft in this session and checked the saved JSON outputs.
Download the actual outputs
- Run report
- Sources, snapshots and original evidence (ZIP)
- Executed command ledger
- Draft SKILL.md
- Evaluation inputs and expectations
- Missing-field output
- Output provenance
Recorded checks and observations
- Observed: SKILL.md passed the local structural validator.
- Observed: Three saved examples cover named owners, missing fields and an unrelated recipe.
- Observed: Four deterministic checks passed, including exact source quotes and unknown missing fields.
Scope, limitations and pinned sources
- The agent read the pinned upstream instructions and used their prescribed tools. Native Claude skill loading and general cross-agent compatibility were not tested.
- Author-applied outputs: the author knew the expected answers and had the requirements in context.
- Fresh-context skill selection, independent review and a without-skill control were not run.
- Saved-output checks do not establish improved behavior or general trigger accuracy.
- The executor was Codex in this conversation; the exact model version was not exposed.
- These examples do not represent independent human users, customer outcomes or a compatibility certification.
Source and evidence
The upstream instructions informed this guide. The exercise and acceptance checks are an OpxSkills test recipe. Where a run is recorded, its actual outputs, environment and limits appear separately above. A recorded example does not establish general model compatibility or a success rate.
Source reviewed · Upstream revision 8a1541c4a3ff
What this skill does
Skill Creator turns a sentence like "draft release notes from merged PRs" into a loadable agent skill.
A skill is a folder with a SKILL.md file. The YAML description is the trigger Claude reads before loading the rest. This skill writes that trigger, the body, and optional script stubs so you are not starting from a blank file.
After it drafts the skill, test the trigger in a fresh session and tighten the description until it fires only on the intended task.
Intended uses
- Bootstrap a personal workflow skill in ~/.claude/skills
- Turn a team runbook into a committed .claude/skills package
- Rewrite a messy prompt into a reusable SKILL.md
Implementation notes
- Put the situation in the description, not the implementation.
- One job per skill — combined skills fail to trigger cleanly.
- Test with the words a teammate would actually type.
Try it, then tell us how it went
A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.