create-agent-skills_skill

This skill helps you verify Claude Code Skill content accuracy by auditing SKILL.md and references, ensuring claims stay current.

TypeScript

10.5k

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 everyinc/compound-engineering-plugin --skill create-agent-skills

SKILL.md9.3 KB

Overview

This skill provides expert guidance for creating, writing, and refining Claude Code Skills, focused on structure, accuracy, and maintainability. It helps authors audit skill content, extract verifiable claims, and produce a clear verification report with suggested updates and schedules. Use it to ensure skills remain accurate as external tools and services evolve.

How this skill works

The workflow walks an auditor through selecting a skill, reading all skill files, categorizing the skill by dependency type, and extracting verifiable claims (CLI flags, API endpoints, framework patterns, file paths). It then verifies claims using the appropriate methods (local commands for CLI, Context7 for libraries/frameworks, web searches for integrations) and generates a freshness report with recommended fixes and a re-verification cadence.

When to use it

Before publishing a new skill to ensure claims are accurate and sources cited
When updating an existing skill after tool, API, or framework releases
As part of a periodic maintenance cadence to prevent stale instructions
When triaging user bug reports that might stem from outdated external dependencies
During code review to confirm documented commands and file paths match reality

Best practices

Always read the full SKILL.md, workflows/, and references/ before verifying
Categorize the skill by its primary dependency to choose the right verification method
Extract concrete, testable claims (tool names, flags, endpoints, paths) for verification
Use local checks for CLI tools and Context7/web searches for APIs and frameworks
Include a verification date and recommended re-check interval in the report

Example use cases

Audit a skill that documents a third-party API to detect deprecated endpoints or auth changes
Verify CLI commands and flags in a build-oriented skill by running version and help checks
Check framework-specific code patterns against the latest library docs via Context7
Produce a step-by-step freshness report for maintainers after a major service release
Automate periodic verification reminders based on the suggested cadence

FAQ

Extract every concrete claim that could become incorrect: tool names and flags, endpoints and auth methods, framework APIs, and file path expectations. Report the total count so reviewers know scope.

What if Context7 or websearch returns ambiguous results?

Flag those claims as 'Could Not Verify' in the report, note the ambiguity, and recommend manual confirmation or contacting the service owner.