Template Optimizer
Documentstemplate-optimizerTemplate Optimizer is an Agent Skill for documents published in the yifanfeng97/Hyper-Extract repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Optimize YAML templates for Hyper-Extract. Use when: "optimize template", "fix YAML issues", "improve quality", "lint template" Trigger: After creating templates or during review Skip: Creating new templates (use brainstorm + designer instead)
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/yifanfeng97/Hyper-Extract/blob/HEAD/hyperextract-skills/template-optimizer/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/template-optimizer, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Template Optimizer
Automatically analyze and optimize YAML templates by applying best practices and fixing common issues.
Workflow
1. Parse YAML → 2. Analyze Issues → 3. Apply Fixes → 4. Generate Report
Step 1: Parse YAML
Load the YAML template and validate basic structure.
Step 2: Analyze Issues
Check against these rules:
| Rule | What to Check |
|---|---|
| rules-naming.md | Field naming standardization |
| rules-multilingual.md | Language consistency |
| rules-field-count.md | Information density |
| rules-consistency.md | Schema vs Guideline separation |
| rules-hypergraph-grouping.md | Hypergraph grouping strategy |
Step 3: Apply Fixes
Apply fixes based on optimization level:
| Level | Description | Example |
|---|---|---|
| Auto-fix | Safe changes that always improve | relation_type → type |
| Suggest | Changes that may need review | Field count > 5 |
| Review | Design decisions | Relation type openness |
Step 4: Generate Report
Output changes with explanations for learning.
Detection Rules
Rule 1: Multi-language Consistency
❌ Pattern: [a-zA-Z]+\([^)]+\) in zh fields
❌ Pattern: Chinese chars in en fields
Fix: Separate language content, use pure Chinese in zh, pure English in en.
Rule 2: Field Naming
❌ relation_type → type
❌ event_date → time
❌ entity_type → type
Fix: Standardize to concise names.
Rule 3: Field Count
⚠️ entities.fields > 5 → Flag for review
⚠️ relations.fields > 5 → Flag for review
Fix: Simplify to essential fields, use priority: Essential → Important → Optional.
Rule 4: Schema-Guideline Separation
❌ Repetition of field definitions in guideline
❌ Schema descriptions in rules
Fix: Schema defines WHAT, Guideline defines HOW TO DO WELL.
Rule 5: Hypergraph Grouping
❌ relation_members: participants (string) + role field exists
❌ relation_id contains participant field
Fix: Use nested grouping relation_members: [group_a, group_b] when entities partition by roles.
Design Principles
Schema vs Guideline
| Schema Defines | Guideline Defines |
|---|---|
| Field names | Extraction strategy |
| Field types | Quality requirements |
| Field descriptions | Creation conditions |
| Required/optional | Common mistakes |
❌ Wrong: Guideline repeats schema definitions ✅ Correct: Guideline explains how to extract well
Information Density
For entities and relations:
| Field Priority | Examples |
|---|---|
| Essential | source, target, participants |
| Important | type, time, location |
| Optional | description, metadata |
Max: 5 fields per component
Naming Conventions
| For | Use | Example |
|---|---|---|
| Template name | CamelCase | EarningsSummary |
| Field names | snake_case | company_name |
| Tags | lowercase | finance, investor |
Reference Files
| Topic | When to Check |
|---|---|
| rules-naming.md | Field naming issues |
| rules-multilingual.md | Language consistency |
| rules-field-count.md | Too many fields |
| rules-consistency.md | Schema-guideline overlap |
| rules-hypergraph-grouping.md | Role field vs nested grouping |
Output Format
# Optimization Report
## Changes Made
| File | Issue | Fix | Level |
|------|-------|-----|-------|
| template.yaml | `relation_type` | renamed to `type` | Auto-fix |
| template.yaml | Mixed language | Fixed zh/en separation | Auto-fix |
| template.yaml | 7 relation fields | Suggested simplification | Suggest |
## Summary
- Auto-fix: 2
- Suggestions: 1
- Manual review: 0
Integration
Full workflow: brainstorm → designer → optimizer → validator
↓
Apply best practices
Auto-fix common issues
When to use:
- After creating new templates
- Before validator
- During template review
- Batch optimization of existing templates