Urgent.News

What's breaking now, across thousands of outlets.

AI

Build and Test an AI Agent Skill with SKILL.md and Python

AI agents are good at interpreting intent, but they are not a reliable place to hide every rule in a workflow. If an agent must count characters, parse a file, or refuse to overwrite an existing artifact, prose instructions alone make the result harder to verify. An Agent Skill gives that workflow a reusable home. The skill describes when it should activate and how the agent should reason. Small…

AI agents are skilled at interpreting intent, but they are not ideal for embedding every rule within a workflow. Counting characters, parsing files, or preventing overwriting existing artifacts can be challenging to verify using only prose instructions. An Agent Skill provides a reusable home for such workflows. The skill outlines when it should activate and how the agent should reason, while minor scripts perform repeatable checks.

Tests safeguard the behavior when the skill evolves. This guide demonstrates how to construct a compact commit-crafter skill using SKILL.md and Python, based on the public how-to-create-a-skill-tutorial repository. The skill is licensed under MIT and documents the open Agent Skills specification.

Key points:

1. Create a skill structure with a SKILL.md file and a scripts/ directory containing validation scripts.

2. Keep long references in the references/ directory and test scripts independently.

3. Write a triggerable SKILL.md that describes the skill's purpose and activation phrases.

4. Implement a deterministic script (e.g., check_message.py) to validate Conventional Commit messages, ensuring the model decides and the script verifies.

5. Test the script thoroughly before installing the skill, covering valid messages, unknown types, subject limits, breaking changes, body formatting, and warning behavior.

6. Test the skill with realistic prompts to ensure proper activation and refinement the description as needed.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

The More Powerful the AI, the More the Architecture Matters

The boundaries I designed, the gaps I haven't solved, and why the difference matters Part 14 findings of an experiment: building an LLM-powered support agent with deterministic boundaries.

  • AI architecture crucial for effectiveness, as shown in experiment
  • Deterministic boundaries, allowlist, and gate prevent unexpected dependencies
  • Gate ensures low-risk tasks handled by AI, high-risk tasks require human oversight

Stage 3 of enterprise AI adoption: agents that can read

This is part 3 of a series that started with the full reference diagram in part 0 , covered identity and the gateway in part 1 , and covered data and retrieval in part 2 .

  • Stage 3 of enterprise AI adoption introduces autonomous agents
  • Agent runtime enables independent task execution with containment
  • Read connectors grant role-specific access to internal systems

More from Saturday 19 September →