agentic-principles_skill

Overview

This skill codifies core principles for collaborating with AI agents during software development. It presents practical rules—divide and conquer, context management, abstraction level selection, automation escalation, planning vs execution, and verification—to make agent-assisted work predictable and safe. Use it to standardize workflows and reduce context drift when multiple agents or sessions are involved.

How this skill works

The skill inspects development tasks and maps them to agent-friendly patterns: break large tasks into independent steps, keep conversation context fresh and purpose-specific, and choose the right abstraction mode (high-level vs deep-dive). It recommends when to automate repetitive steps, how to orchestrate multiple agents with clear roles, and concrete verification checkpoints like tests and draft PRs.

When to use it

• At the start of any AI agent session to set collaboration rules
• Before tackling complex or cross-file changes to define plan vs execution
• When establishing context management strategy for long or multi-topic work
• During onboarding to teach team members agent collaboration patterns
• When repetitive manual steps appear and automation should be considered

Best practices

• Always split large tasks into small, verifiable stages with independent checks
• Keep each conversation single-purpose; create HANDOFF documents when sessions grow long
• Select abstraction level deliberately: use high-level for design, deep-dive for debugging
• Automate tasks repeated three or more times and evolve automation incrementally
• Prefer planning mode for risky or large changes; use execution mode in safe, contained environments

Example use cases

• Build a new feature by staging: schema → core logic → API → tests → integration
• Run multi-agent workflow: planning agent defines tasks, analysis agent researches, executor runs builds
• Onboard a teammate by demonstrating single-purpose tabs and HANDOFF.md state transfers
• Convert repeated manual code transformations into a scripted workflow or skill
• Perform production changes by planning thoroughly, generating draft PRs, and running self-verification