3.6 KiB

Raw Blame History

Planner Agent Skill Section Example

Real-world example from planner.md agent showing effective skill integration.

Original (Before Streamlining)

## Relevant Skills

You have access to these specialized skills when planning architecture:

- **agent-output-formats**: Standardized output formats for agent responses
- **architecture-patterns**: System design, ADRs, design patterns, scalability patterns
- **project-management**: Project scope, goal alignment, constraint checking
- **database-design**: Schema design, normalization, query patterns
- **api-design**: API design patterns, endpoint structure, versioning
- **file-organization**: Project structure standards and organization
- **testing-guide**: Testing strategy patterns and coverage approaches
- **python-standards**: Language conventions affecting architecture decisions
- **security-patterns**: Security architecture and threat modeling

When planning a feature, consult the relevant skills to ensure your architecture follows best practices and patterns.

Token Count: ~190 tokens

Streamlined (After Streamlining)

## Relevant Skills

You have access to these specialized skills when planning architecture:

- **architecture-patterns**: Apply for system design and scalability decisions
- **api-design**: Follow for endpoint structure and versioning
- **database-design**: Use for schema planning and normalization
- **testing-guide**: Reference for test strategy planning
- **security-patterns**: Consult for security architecture

Consult the skill-integration-templates skill for formatting guidance.

Token Count: ~90 tokens

Token Savings: 100 tokens (52% reduction)

Key Improvements

Reduced skill count from 9 to 5 (kept only most relevant)
Concise descriptions - One line per skill instead of multi-word descriptions
Action verbs - "Apply", "Follow", "Use", "Reference", "Consult" match planning context
Meta-skill reference - Points to skill-integration-templates for formatting guidance
Removed redundant closing - "When planning..." was verbose

Why This Works

Progressive Disclosure

Full skill content loads on-demand when skills are referenced
Lightweight metadata (skill names) stays in context
Planner can still access all 9 original skills if needed

Token Efficiency

190 tokens → 90 tokens (100 token savings)
Essential skills still listed
No loss of functionality

Maintained Quality

Planner still knows which skills are available
Action verbs guide when to use each skill
Closing sentence provides formatting guidance

Usage in planner.md

Location: plugins/autonomous-dev/agents/planner.md

Full Context:

---
name: planner
description: Architecture planning and design for complex features
model: opus
tools: [Read, Grep, Glob]
---

You are the **planner** agent.

## Your Mission

Design detailed, actionable architecture plans for requested features...

[agent-specific mission and workflow]

## Relevant Skills

You have access to these specialized skills when planning architecture:

- **architecture-patterns**: Apply for system design and scalability decisions
- **api-design**: Follow for endpoint structure and versioning
- **database-design**: Use for schema planning and normalization
- **testing-guide**: Reference for test strategy planning
- **security-patterns**: Consult for security architecture

Consult the skill-integration-templates skill for formatting guidance.

[rest of agent prompt]

implementer-skill-section.md - Implementer agent example
minimal-skill-reference.md - Minimal reference pattern

3.6 KiB Raw Blame History