---
title: "brace"
description: "Design and review AI agent plugins for cross-harness coherence. Checks artifact dependencies between skills, knowledge-layer separation, internal duplication, and whether shared resources actually load on every harness t"
canonical: https://agentpluginsdirectory.com/plugins/brace
last-updated: 2026-10-21
---

# brace
Design and review AI agent plugins for cross-harness coherence. Checks artifact dependencies between skills, knowledge-layer separation, internal duplication, and whether shared resources actually load on every harness the plugin claims to support.
- Slug: brace
- Publisher: Bo Frese
- Repository: https://github.com/bofrese/brace
- Manifest: plugin.json
- Version: 0.1.0
- License: Apache-2.0
- Category (editorial): agent-tooling
- Skills: 4 (harness-verify, plugin-design, plugin-enable, plugin-review)
- MCP servers: 0
- Stars: 0
- Repository created: 2026-09-02
- Repository last pushed: 2026-09-02
- Publisher type: User
- Listing: https://agentpluginsdirectory.com/plugins/brace
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What brace does, in the publisher's words

Build AI agent plugins that hold together, and enable them per project across Claude Code, Codex, and GitHub Copilot without hand-editing four config formats.

A brace is the member that holds a frame square. That is the job: a plugin's skills kept in a coherent shape, and an honest answer about what each harness will actually load.

Two halves, one repo:

- Authoring and review skills: plugin-design and plugin-review check that a plugin's skills form a coherent workflow: that artifacts flow, that knowledge sits in the right layer, that nothing is duplicated or contradicted, that shared resources actually load on every harness the plugin claims, and that the result is publishable to a marketplace rather than only loadable from a checkout. harness-verify re-verifies the harness facts all of that rests on.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/bofrese/brace/HEAD/README.md

## Skills

- harness-verify: Re-verify what every AI agent harness currently does with plugins, against primary vendor documentation, and bring this repo's facts, adapters and skills back in line with it. Use when harness facts feel dated, before claiming cross-harness support to a client, when a vendor announces a plugin or s…
- plugin-design: Design a new AI agent plugin, or restructure an existing one, so its skills form a coherent workflow across agent harnesses. Use when starting a plugin from scratch, when turning a pile of prompts or skills into something structured, when deciding how skills should hand work to each other, or when…
- plugin-enable: Enable or disable AI agent plugins for a project across Claude Code, Codex, and GitHub Copilot, and diagnose what each harness is actually loading. Use when a plugin's skills are not showing up in an agent, when setting up a project to use a plugin checkout, when edits to a plugin are not taking ef…
- plugin-review: Review an AI agent plugin for coherence, whether its skills actually work together as a workflow. Use when asked to review, audit, or check a plugin or a set of skills; when a plugin has grown and nobody is sure the pieces still fit; before publishing or handing a plugin to someone else; or when s…

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.