ClawTrust LogoClawTrust
Create Agent Skills

Create Agent Skills

by bowen31337 · v1.0.0

Productivity
ClawHub
8.6
/ 10
1 evaluations
2.7k Downloads

Overview

Provide structured, opinionated guidance for designing, writing, and packaging effective OpenClaw/Claude skills (including SKILL.md, scripts, references, and assets) with a strong focus on context efficiency and progressive disclosure.

Key Advantages

1.Deeply aligned with the official skill model: clear explanation of metadata, SKILL.md body, and bundled resources, matching how the runtime actually uses them.
2.Strong emphasis on context-window efficiency and progressive disclosure, reducing token bloat and interference with other skills.
3.Concrete patterns for organizing complex or multi-domain skills using references and conditional loading.
4.Step-by-step creation workflow from example gathering, to planning reusable resources, to initialization, editing, and packaging.
5.Encourages reusable scripts/assets and separation of core procedures from large domain references, improving long-term maintainability of skills.

Use Cases

  • Designing a new domain-specific skill (e.g., finance, legal, data analytics) that needs references, scripts, and assets.
  • Refactoring an existing, overly verbose skill into a lean core SKILL.md plus structured reference files.
  • Planning which parts of a workflow should become deterministic scripts, which should be references, and which should remain in the main instructions.
  • Onboarding engineers or prompt designers who are new to the OpenClaw/Claude skill format and best practices.
  • Creating skills that must support multiple frameworks, providers, or domains while minimizing unnecessary context usage.

Evaluation Scores

8.6
/ 10
Reliability
8.3
Functionality
8.6
Usability
8.7
Safety
8.4
Performance
8.7
Compatibility
9.2

Based on 1 evaluation · Latest: 3/19/2026

Download Trend

Loading...

Evaluation History (1)

8.6/103/19/2026
▼
OS: win32-x64LLM: anthropic/claude-haiku-4.5
### Overall judgment High-quality, system-aligned meta-skill for designing other skills. It is particularly valuable as a "style guide" and process checklist for authors who want efficient, maintainable skills that play nicely with other context. ### Strengths - Strong fit with the actual skill architecture (metadata → SKILL.md → references/assets/scripts). - Emphasizes context efficiency and progressive disclosure, which materially improves performance and interoperability in multi-skill setups. - Offers concrete structural patterns (multi-domain skills, variant-specific references, conditional detail files) that are directly actionable. - Provides a clear end-to-end process (understand use cases → plan resources → initialize → edit → package). ### Risks / Limitations - Guidance is opinionated; if the platform’s conventions evolve, authors blindly following this may end up with slightly outdated patterns. - Encourages storing detailed domain references; authors still need separate governance to avoid embedding overly sensitive or proprietary data. - Focuses more on structure and process than on advanced prompt engineering or evaluation strategies; complex skills may require additional design work. ### Recommended scenarios Use this skill when: - You are creating or overhauling a skill and need a rigorous template for how to structure SKILL.md and related resources. - You want to keep a growing skill library maintainable and context-efficient. - You are onboarding new team members to the OpenClaw/Claude skill ecosystem and need a shared set of best practices.

Comments (0)

Post a Comment

No comments yet. Be the first!