Agent Skills

2026-05-27 · Source: AI & ML – Radar · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Intermediate, long

Summary

The "Agent Skills" project, with over 27K stars, provides a framework to embed senior engineering practices into AI coding agents. It addresses agents' default tendency to skip crucial Software Development Lifecycle (SDLC) steps like writing specs, tests, and conducting reviews. The system uses "skills," which are Markdown files containing actionable workflows and checkpoints, injected into the agent's context. These skills are organized across six SDLC phases and include features like anti-rationalization tables, non-negotiable verification, progressive disclosure, and strict scope discipline. The project intentionally aligns with Google's established engineering practices to ensure agents produce reliable, reviewable code.

Key takeaway

For AI Engineers integrating coding agents, you must explicitly encode senior engineering discipline into agent workflows to prevent shortcuts and ensure reliable software delivery. Implement structured "skills" and anti-rationalization tables to enforce critical SDLC steps like specification, testing, and scope discipline. This approach mitigates the risk of unmanageable code changes and increased incidents, fostering a more robust development process.

Key insights

AI coding agents require explicit workflow-based "skills" to enforce senior engineering practices and prevent shortcuts.

Principles

• Process over prose: workflows are actionable.
• Anti-rationalization tables preempt agent excuses.
• Verification is nonnegotiable; tasks need evidence.

Method

Agent Skills injects Markdown-based "skills" (workflows with checkpoints and exit criteria) into an agent's context, activated by a meta-skill router based on the SDLC phase.

In practice

• Install Agent Skills via marketplace.
• Adopt anti-rationalization tables for teams.
• Convert internal docs to workflows.

Topics

• AI Coding Agents
• Software Development Lifecycle
• Engineering Practices
• Agent Harness Engineering
• Code Review
• Test-Driven Development

Code references

• addyosmani/agent-skills

Best for: AI Engineer, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Agentic Engineering — Effective agentic engineering requires structured workflows to plan, steer, and verify AI-generated code.
• Stop giving agents the whole computer — AI coding agents are advancing in autonomy and capability, but demand stricter security and structured development workflows.
• Evals Skills for Coding Agents — Improving infrastructure around AI agents, especially evaluation capabilities, is more critical than solely improving the underlying model.
• Coding Agents: AI Driven Dev Conference — The event offers practical, production-tested strategies for AI-assisted code development and agentic engineering.
• A Year Of Vibes — Agentic coding tools are fundamentally reshaping software engineering workflows, challenging traditional practices and human-machine relationships.
• Level-up your AI Agent with Skills Engineering — Skills engineering explicitly guides AI agent decision-making through detailed instructions, improving reliability and performance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI & ML – Radar.