skill-load-tips

Refactor SKILL.md files with required reading routers and STOP directives.

182|8|Updated Apr 4, 2026
One-click install
npx skills add https://github.com/compozy/agh --skill skill-load-tips
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-load-tips
Source: https://github.com/compozy/agh/tree/main/.agents/skills/skill-load-tips
Command: npx skills add https://github.com/compozy/agh --skill skill-load-tips

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It fixes skills where the agent mostly reads only the SKILL.md body and ignores the load-bearing content in references, leading to shallow, inconsistent, or incorrect results.

Core Features & Use Cases

  • Required Reading Router: Adds a task-to-file mapping so the agent knows exactly which reference files must be read for each task type.
  • Hard STOP directives: Replaces soft prompts with imperative “STOP. Read … in full” instructions tied to specific operating steps.
  • Gist tripwire trimming: Shrinks duplicated inline checklists into brief tripwires while forcing the agent to load the full reference instead.
  • One-level reference flattening: Ensures references link directly from SKILL.md (not from other reference files) and adds a top “Contents” section for large reference files.
  • Use Case: Audit a long-running production skill that “feels right” in the body but underperforms because the agent never fans out into references; refactor it so reference files are reliably loaded and used as the source of truth.

Quick Start

Ask an agent to audit the target skill by diagnosing which reference files are likely never loaded and then refactor its SKILL.md by adding a router, STOP directives, and trimmed tripwire inline content.

Frequently Asked Questions about skill-load-tips

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
Why does my agent ignore reference files and only read the SKILL.md body?

Agents exhibit body-first reading behavior and skip references due to soft phrasing or nested link structures. Refactoring the SKILL.md with a required reading router and hard STOP directives forces the agent to load those files.

How do I ensure an agent loads context from YAML frontmatter references?

To ensure context loading from references, implement one-level reference flattening so links point directly from the SKILL.md. Adding a top Contents section helps the agent navigate large reference files without dropping critical scope.

What is the best way to refactor agent instructions for prompt reliability?

The best way to refactor agent instructions for prompt reliability is replacing soft prompts with imperative STOP directives tied to specific operating steps. This enforces deep reads of the operational truth contained in reference files.

Can I trim inline checklists in SKILL.md without losing critical scope?

Yes, you can trim duplicated inline checklists into brief gist tripwires. This shrinks the SKILL.md body while forcing the agent to load the full reference file, ensuring no critical scope is lost during partial reads.

When do I need to add a required reading router to a skill?

You need a required reading router when a long-running production skill underperforms because the agent never fans out into references. The router provides explicit task-to-file mapping so the agent knows exactly which reference files to read.