lbbniu-skill-creator_skill

This skill guides you through crafting effective Claude skills, including structure, references, and progressive disclosure for efficient context usage.
  • Python

2.5k

GitHub Stars

3

Bundled Files

2 months ago

Catalog Refreshed

3 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill lbbniu-skill-creator

  • _meta.json463 B
  • LICENSE.txt11.1 KB
  • SKILL.md17.7 KB

Overview

This skill is a practical guide for designing, building, and packaging skills that extend Claude with domain-specific knowledge, workflows, and tool integrations. It focuses on creating compact, reusable skill packages that maximize usefulness while minimizing context cost. Use it to plan what to include, how to organize resources, and how to iterate from examples to production-ready skill bundles.

How this skill works

The skill inspects a proposed capability and recommends a minimal, staged structure of metadata, core instructions, and optional bundled resources (scripts, reference docs, and output assets). It prescribes progressive disclosure so only the right amount of context is loaded when needed. It also provides a stepwise workflow: understand use cases, plan reusable contents, initialize a template, implement resources, package, and iterate based on real usage.

When to use it

  • You are creating a new skill to add domain knowledge, workflows, or API integrations to Claude.
  • You are updating an existing skill and need guidance on trimming context or reorganizing resources.
  • You want a repeatable process for packaging scripts, references, and output assets for deterministic tasks.
  • You need to decide what to keep in core instructions vs. what to store as references or assets.
  • You want to design a skill that triggers only for relevant user queries and stays token-efficient.

Best practices

  • Keep core instructions concise—only include information Claude truly needs up front.
  • Use progressive disclosure: small metadata + core instructions, then load references or scripts on demand.
  • Match freedom level to task fragility: high freedom for heuristics, low freedom for fragile sequences.
  • Bundle deterministic code as executable scripts and store large domain docs as separate reference files.
  • Avoid extraneous files or duplicate content; reference external docs rather than copy them into the skill.
  • Organize references by domain or variant so Claude can load only the relevant file.

Example use cases

  • Build a PDF-processing skill with a small core workflow and a scripts/rotate_pdf.py for deterministic rotation.
  • Create a finance skill that stores table schemas in references/ and exposes a compact query workflow in core instructions.
  • Package a frontend-boilerplate skill where assets/ contains template projects and core instructions explain customization points.
  • Refactor a large monolithic skill by moving large examples and deep reference material into separate files referenced from the core.
  • Design a conditional-loading skill that presents basic steps first and loads advanced guides only when requested.

FAQ

Keep only the procedural essentials and trigger criteria in the core; move detailed schemas, long examples, and API specs to reference files that load on demand.

When should I include executable scripts?

Include scripts when repeated deterministic code is needed or when token efficiency and reproducibility are priorities.

How should I handle multiple variants (frameworks or providers)?

Provide a short selector in the core and store variant-specific details in separate reference files so Claude only loads what’s relevant.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational