What does this skill do, and when should you use it?
This is a meta-skill: its entire purpose is to guide Claude in creating new Agent Skills for you. It ships a complete creation workflow — reading bundled docs/ reference material, questioning the user to define scope, scaffolding the directory, writing a SKILL.md with YAML frontmatter, adding examples and supporting files, then testing and committing to git. It distills the core architecture principles (progressive disclosure, one capability per skill, descriptions that state both what and when) and provides four full end-to-end examples. It is distributed as part of the disler/claude-code-hooks-multi-agent-observability repository's bundled skill collection.
- Reads three required documentation files in the skill's bundled docs/ directory covering skill architecture and design principles
- Asks the user clarifying questions to define the new skill's purpose, triggers, and required resources
- Creates .claude/skills/<skill-name>/ and writes a SKILL.md with name/description frontmatter
- Writes instructions and 2-4 concrete examples, optionally adding scripts (with PEP 723 inline dependencies), templates, and reference files
- Demonstrates extending an existing skill (e.g., appending a spell-correction mapping file to a transcription skill)
- Guides committing the skill to git so teams receive it automatically via pull
- An engineer who wants to codify recurring work like 'Python code review' into a skill Claude triggers automatically
- A data analyst who needs a CSV analysis skill with pandas/matplotlib scripts and wants it packaged correctly
- A technical writer creating a multi-file documentation skill structured around a company style guide (templates, guidelines, examples)
- A user with an existing skill who wants to extend it safely without breaking its structure
- A newcomer to Agent Skills who wants to study the spec and anti-patterns before writing their first skill
- Users who just want ready-made capabilities and don't intend to author skills — this skill produces new skill files, not direct business results
- Scenarios needing the creation flow to run autonomously outside an agent host (e.g., a bare model API) — the workflow's file and directory operations require filesystem access from the caller
How do you install this skill?
- Shell / CLI
- Local filesystem
docs/ reference files bundled with the skillgit for team sharing (optional)
The skill ships inside the repository's .claude directory. Per the README, copying the whole .claude directory to your project root makes the project skills (including this one) available:
Claude Code
git clone https://github.com/disler/claude-code-hooks-multi-agent-observability.git
cp -R claude-code-hooks-multi-agent-observability/.claude /path/to/your/project/Note: the repo does not provide a command to install this single skill in isolation; if you only want the meta-skill, manually copy .claude/skills/meta-skill/ into your project's .claude/skills/.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Create a skill that reviews Python code for PEP 8 compliance, security issues, and performance
- Build a skill for analyzing CSV data with statistics and visualizations, outputting a markdown report
- Create a skill for writing technical documentation following our company style guide, supporting API docs and user guides
- Add spell correction to our existing transcribe skill
The skill triggers via its description: when the user wants to create a new skill, extend Claude's capabilities, or package domain expertise into a reusable skill. The flow is: read the three docs/ files → ask the user scoping questions → create .claude/skills/<skill-name>/ → write SKILL.md (name max 64 chars; description max 1024 chars covering both what it does and when to use it) → write instructions and 2-4 examples → optionally add supporting files (scripts need chmod +x and PEP 723 dependency headers) → test triggering and execution → commit to git for team sharing. Personal-only skills belong in ~/.claude/skills/ instead of the project directory.
What are this skill's strengths and limitations?
- Bundles complete reference docs and four end-to-end examples, covering zero-to-git-commit
- Enforces best practices: progressive disclosure, single-responsibility skills, descriptions that state triggers
- Covers creation patterns for three skill shapes: instruction-only, script-backed, and multi-file
- The skill file is its own documentation; teams get it via a simple git pull
- Output quality depends on the bundled docs/ files — you must copy them too, not just SKILL.md
- Repository license is not stated; confirm before commercial use
- Script dependencies in the examples (pandas, matplotlib, etc.) require the user to set up a Python environment
- It only teaches Claude Code-style skill authoring and does not cover packaging differences on other Agent Skills platforms
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Create New Skills (Meta-Skill) this page | 62 · Recommended | ★ 1.5k | 8mo ago | — |
| Skill Creator | 70 · Strongly recommended | ★ 2.2k | 12d ago | MIT |
| Claude Code Plugin Skill Development ✓ Anthropic · Official | 51 · Use with care | ★ 38k | 11d ago | Apache-2.0 |
| Skill Creator | 48 · Use with care | ★ 3.1k | 8mo ago | MIT |
| DeepTutor Skill Creator | 57 · Use with care | ★ 41k | 17d ago | Apache-2.0 |
The same repository bundles other skills (e.g., transcription, morning debrief); this one is the meta-skill — it performs no business task itself but teaches Claude to manufacture other skills. If you only need ready-made capability, pick one of those instead.
How did FollowSkills review this skill?
A meta-skill for creating new skills, requiring reading official docs on the three-level progressive-disclosure structure before starting — pure documentation generation.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
This review was completed under FSRS 1.0. Dimension scores are not converted; it is queued for re-review under FSRS 2.0.
See the full review method →