Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
The author's original description, preserved in English. It describes intent, not verified performance.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
developmentreferencesscripts
01 / OVERVIEW
Research summary
Human reviewed
Skill Atlas editorial baseline · editorial-0.1.0
A process skill for drafting, testing, revising and evaluating other Skills. Its main value is the feedback loop: write a draft, create prompts, inspect results, then rewrite based on evidence. It does not write a good Skill for you; it structures the work and asks for repeated validation.
Review scope
Deterministic static analysis of collected SKILL.md and supporting text/code. Machine-assisted candidates and findings await human editorial review; no dynamic behavior or effect is assessed.
Observable capabilities
Guides a creator from intent to draft, test prompts, review and revision.
Separates qualitative reviewer feedback from quantitative evaluation ideas.
References helper scripts and an evaluation viewer for reviewing multiple runs.
Handles both new Skills and improvements to existing ones.
Limitations & scope
The instructions assume access to a model runner and local helper scripts; without them the evaluation loop cannot be executed.
Test prompts and success criteria still need human design; the Skill does not know your product or users.
It contains long sections and many references, so activation cost can be noticeably higher than a short instruction-only Skill.
Relevant for
Skill authors
Teams maintaining internal Skill libraries
Reviewers who need a repeatable revision workflow
02 / STRUCTURE
What ships with the skill
The file inventory comes from the collected source at the recorded commit. Paths and sizes are structural observations, not a measure of quality.
{
"name": "skill-creator",
"description": "Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy."
}
03 / WORKFLOW
A reading of the operating sequence
This workflow reconstructs the apparent sequence from source material. Unless controlled evaluation is explicitly recorded, it is provisional—not an observed trace.
01
Input
Confirm scope
Read the trigger conditions and decide whether this skill matches the current task.
Static findings identify points to inspect before installation. They do not establish that a skill is safe, malicious, or effective.
Items warrant review
328 deterministic lexical finding(s) require review across the collected text. Findings can occur in documentation and do not prove execution or exploitability; binary-only files were not scanned.
Network accessObserved
File writesObserved
Credential useNot observed
Note
File modification is described or implemented
312 occurrences
This line contains a possible file-write or redirection operation. Check destination paths and overwrite behavior; the static review did not write any file. This rule matched 312 location(s) in the collected text; 3 representative location(s) are cited. Matching text may be documentation, and the analyzer did not execute it.
A URL or network client appears in this source line. Review destinations, data sent, and trust boundaries; static text cannot establish actual network activity. This rule matched 16 location(s) in the collected text; 3 representative location(s) are cited. Matching text may be documentation, and the analyzer did not execute it.
The Skill references executable evaluation helpers. Inspect those scripts before running them; this record does not verify their behavior.
05 / CONTEXT
Where context can be spent
Discovery, activation, and resource layers have different meanings. Exact values appear only when the official token-counting method was run with a named model.
Exact token counts unavailable
All token values, including per-file inventory, are null until the pipeline performs an official exact count with an explicit model.
No token measurement was recorded with the required official method. This dossier does not substitute character counts, file sizes, or estimates.
06 / EVALUATION
Controlled evaluation
No controlled evaluation
This skill has only the evidence identified above. No task success, token efficiency, latency, or quality advantage is claimed.
Open the immutable source at https://github.com/anthropics/skills/tree/e4bb5d960d87126ff3b3a8af1a104b7db67560ba/skills/skill-creator/SKILL.md.
Review the license, prerequisites and all bundled resources.
Install the skill into a personal or project-level Claude Code skills directory.
Restart or refresh Claude Code so the skill can be discovered.
Verify installation
Open Claude Code and confirm the skill appears in the installed skills list, then run a representative prompt below.
Prerequisites
The instructions assume access to a model runner and local helper scripts; without them the evaluation loop cannot be executed.
Test prompts and success criteria still need human design; the Skill does not know your product or users.
It contains long sections and many references, so activation cost can be noticeably higher than a short instruction-only Skill.
Example prompts
Representative task
Use the skill-creator skill to help with this task. First summarize what the skill will do, then execute its documented workflow and show the source-linked evidence for each major decision.
Inspect before use
Read the skill-creator skill files at https://github.com/anthropics/skills/tree/e4bb5d960d87126ff3b3a8af1a104b7db67560ba/skills/skill-creator/SKILL.md. Explain its workflow, required tools, permissions and the main risks before I decide whether to install it.
08 / VERDICT
A decision, with its limits attached
Strengths
Makes evaluation part of the writing process instead of an afterthought.
Give the creator concrete stages and artifacts to inspect.
Considerations
Value depends on actually running test prompts and reviewing outputs.
Not a substitute for clear domain knowledge in the Skill being created.
Research conclusion
Worth reading as a methodology template if you plan to author Skills seriously. Treat it as a process guide, not a one-click generator.
09 / PROVENANCE
Traceable by construction
The dossier is a time-bound reading of a public artifact, not an endorsement. Use the source link to inspect the exact upstream material.
Deterministic lexical parsing, YAML frontmatter validation, file inventory, source-described workflow extraction, and heuristic provisional classification of the supplied files only. Generated by machine; no LLM, execution, dynamic testing, or human editorial approval.