Generated Claude Skill Names are "wrong"

[fgrehm]

As per Claude Docs:

Field requirements:

• name: Must use lowercase letters, numbers, and hyphens only (max 64 characters)
• description: Brief description of what the Skill does and when to use it (max 1024 characters)

But agent-os is generating a "standard_name_humanized_capitalized" for all skills and I have a feeling it's not picking up the skills when it should based on my interactions with it for the past week or so.

Here's what I got for my project for example:

$ head -n 4 .claude/skills/**/*.md | grep 'name:'

name: Backend API
name: Backend Migrations
name: Backend Models
name: Backend Queries
name: Frontend Accessibility
name: Frontend Components
name: Frontend CSS
name: Frontend Responsive
name: Global Coding Style
name: Global Commenting
name: Global Conventions
name: Global Error Handling
name: Global Tech Stack
name: Global Validation
name: Testing Test Writing

I'll try to put up a PR for this today if time permits

[fgrehm]

Update: Fix Implemented + Naming Best Practices Question

I've fixed the immediate bug in my fork (commit):

Problem: Templates use {{standard_name_humanized_capitalized}} in YAML → "Frontend Components" ❌
Fix: Use {{skill_name}} instead → "frontend-components" ✅

This makes names valid per Claude's requirements, but I noticed Claude's best practices actually recommend action-oriented names (gerunds):

• ✅ Recommended: handling-errors, styling-components, managing-dependencies
• 😐 Our output: error-handling, frontend-css, backend-api (noun phrases)

The tension: Standards are topics, Skills are actions.

---

Questions

1. Is kebab-case enough for now? Should I submit a PR with just the format fix, or wait to address action-oriented naming?
2. How important is this? Are other users experiencing skill discovery issues in practice like me, or are noun phrases working fine?

Happy to implement any approach, just want to align before submitting a PR. Thoughts?

[phpmacher]

Personally, I think your idea and its implementation are very good.

I must admit that I don't have enough experience with it to judge whether one or the other would be better, but of course I trust the inventors of the skills first and foremost.

I would be very interested in feedback on this topic and whether @CasJam would like the suggested improvement.

[fgrehm]

FWIW - even with those changes I'm still not sure Claude is picking up the skills after the name changes.

I was playing with claude skills in another project that does not use agent-os and noticed that claude asks for confirmation before using them, but I personally haven't got prompted for that when using agent os commands like /implement-tasks for example (I did double check that those were not allowed in my claude settings already).

I'm doing some more digging around this, but I'm curious to know if others are seeing their skills being used out in the wild.

[joachim-isaksson]

This fixes the issue that agent-os skills don't load out of the box with opencode-skills plugin also, it fails loading them with the message that they don't follow the naming spec.