skill-creator-primer_skill

Overview

This skill provides foundational guidance for authoring agent skills and should be loaded before creating or updating skills. It captures practical rules, templates, and validation steps to produce focused, triggerable skills that fit within shared LLM context limits. Use it as a checklist and reference while designing skill metadata and content.

How this skill works

The skill explains how skills act as context modifiers rather than executable code, detailing how instructions, tool permissions, and optional model changes are injected into conversation context. It describes how the model selects skills via language-based reasoning and teaches techniques to make descriptions reliably trigger. It also provides guidelines for token budgeting, degrees of authorial freedom, and post-creation self-review and validation.

When to use it

• Before starting a new skill authoring session
• When updating an existing skill’s description or behaviour
• To audit token usage and reduce context bloat
• When deciding how prescriptive or flexible a skill should be
• Prior to running automated validation on a skill package

Best practices

• Write concise, comprehensive descriptions so the model can select the skill reliably
• Keep the main skill document short; move long references to external files
• Choose an appropriate degree of freedom (high/medium/low) and add guardrails for fragile flows
• Assertively phrase activation conditions to reduce under-triggering
• Review and thin content after creation: remove duplication and low-value prose
• Validate against the official skills spec and run automated checks before publishing

Example use cases

• Design a skill that performs a multi-step data-processing workflow with tightly ordered steps
• Create a descriptive metadata block that ensures the model activates a dashboard-building skill when dashboards are mentioned
• Audit an existing skill to reduce token footprint and move large reference materials to bundled resources
• Build a template for parameterised scripts where consistent formatting and safe defaults are required