- Home
- Skills
- Basher83
- Agent Auditor
- Skill Creator
skill-creator_skill
- Python
4
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 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 basher83/agent-auditor --skill skill-creator- SKILL.md20.3 KB
Overview
This skill helps authors design and package reusable agent skills that extend an AI assistant with domain knowledge, workflows, and tools. It guides you through a concise manifest, modular resource layout, and progressive-loading patterns that keep context usage efficient. Use it to convert repeatable procedures into small, maintainable skill packages.
How this skill works
The skill inspects the target workflow and recommends a minimal manifest (name + description) that triggers the skill when relevant. It prescribes a three-tier organization: small manifest metadata, a compact instruction document, and optional bundled resources (scripts, references, assets) that are loaded only when needed. It also offers patterns for splitting content, choosing degrees of freedom (high/medium/low), and packaging scripts for deterministic execution.
When to use it
- When you want to create a new skill to teach the assistant a repeatable workflow or domain procedure.
- When packaging code, templates, or assets the assistant should use but not always load into context.
- When the assistant needs access to company-specific schemas, API docs, or policies as references.
- When you need to reduce context bloat and ensure only relevant content is loaded.
- When converting informal instructions into a discoverable, triggerable skill manifest.
Best practices
- Keep the manifest metadata focused: clear name and description determine when the skill triggers.
- Keep the main instruction document concise; move long or variant-specific content into reference files.
- Use scripts for fragile, deterministic steps and load them as executable resources rather than inline text.
- Organize references one level deep and include a short TOC or grep patterns for large files.
- Design progressive disclosure: metadata always available, main instructions loaded on trigger, resources loaded as needed.
Example use cases
- Create a PDF-processing skill with a small workflow and a scripts/rotate_pdf.py for deterministic rotation.
- Package an API-integration skill where references/api_docs.md is loaded only when API calls are required.
- Build a compliance skill that keeps policy text in references/ and example templates in assets/.
- Convert an image-editing workflow into a skill with example prompts, asset templates, and repeatable scripts.
- Organize a multi-provider deployment skill that loads provider-specific reference files only after the user selects a provider.
FAQ
Keep only core workflow guidance and selection heuristics in the main file; move long examples, schemas, and exhaustive references into separate files that are loaded on demand.
When should I include executable scripts?
Include scripts when a step must be deterministic, error-prone, or frequently rewritten. Scripts reduce token cost and can be run without loading their source into the assistant context.