Skip to content

Add Agent Skill for generating new Agent Skills#1061

Closed
kodurusravani34 wants to merge 1 commit intowebinstall:mainfrom
kodurusravani34:add-agent-skill-generator
Closed

Add Agent Skill for generating new Agent Skills#1061
kodurusravani34 wants to merge 1 commit intowebinstall:mainfrom
kodurusravani34:add-agent-skill-generator

Conversation

@kodurusravani34
Copy link
Copy Markdown

Added an Agent Skill that generates new Agent Skills from documentation or repository inputs. Helps simplify and scale skill creation.

@David-Bosnic
Copy link
Copy Markdown

David-Bosnic commented Mar 18, 2026

This has no use case for webi and would just adds a markdown file that pretty much does nothing.
Refer to @detox-24 comment and mine after.
Closing

@detox-24
Copy link
Copy Markdown
Member

Hello @kodurusravani34! Your idea is in the right place. Might I suggest some changes/improvements to this PR.

  1. Wrong file structure

Agent Skills follow the agentskills.io spec, which requires a named directory containing a SKILL.md file, not a standalone .md file. The correct structure looks something close to this:

_skills/agent-skill-generator/
├── SKILL.md
└── references/        (optional but encouraged)
    └── REFERENCE.md

Your file was placed at agent-customization/agent-skill-generator.md which does not follow the spec. Skills in this repo live under _skills/. For a real example of what this looks like in practice, check out PR #1059 by @coolaj86, that is a good reference for structure and depth.

  1. Missing YAML frontmatter

The SKILL.md must start with a YAML frontmatter block. At minimum:

---
name: agent-skill-generator
description: Generates a new Agent Skill for any tool by analyzing its documentation URL or repository. Use when asked to create or scaffold a new Agent Skill.
---
  1. Add a verification step

Issue #1056 specifically calls this out. Once the skill generates a new Agent Skill, there should be a way to confirm the output actually works, not just that it looks right. Even a short checklist would do: does the generated skill have valid frontmatter? Can an agent follow the instructions end to end? Does it cover both agent use and human-via-agent use?

@David-Bosnic, I think this one deserves a chance to update, thoughts?

@David-Bosnic
Copy link
Copy Markdown

David-Bosnic commented Mar 18, 2026

Fair. I did shoot the gun.
@kodurusravani34
Open a new pr with a better description and follow @detox-24 suggestions.
A better understanding of what and why will help in the future when people look back into the PR
even though the preposed merge is small.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants