From agent-config
Initialize a unified LLM coding agent configuration. Creates AGENTS.md as the single source of truth and symlinks it to tool-specific config files. Use when the user wants to set up agent config or unify AI tool settings.
npx claudepluginhub tqer39/claude-code-marketplace --plugin agent-configThis skill uses the workspace's default tool permissions.
Create `AGENTS.md` as the single source of truth for LLM coding agent guidelines. Then symlink it to each tool's config file. This keeps all agents aligned on the same rules with zero duplication.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides MCP server integration in Claude Code plugins via .mcp.json or plugin.json configs for stdio, SSE, HTTP types, enabling external services as tools.
Create AGENTS.md as the single source of truth for LLM coding agent guidelines. Then symlink it to each tool's config file. This keeps all agents aligned on the same rules with zero duplication.
Based on arxiv 2602.11988: concise context files (~32 lines) outperform verbose ones.
Write the following to AGENTS.md in the project root:
# Agent Guidelines
## Principles
- Always prefer simplicity over pathological correctness
- YAGNI: do not add functionality until it is necessary
- KISS: prefer the simplest solution that works
- DRY: avoid duplication of logic
- No backward-compatibility shims or fallback paths unless they come for free without increasing cyclomatic complexity
## Code Style
- Follow existing project conventions; do not introduce new patterns gratuitously
- Prefer standard library solutions over third-party dependencies
- Write small, focused functions with clear names
- Comments explain "why", not "what"
## Changes
- Make the minimal change that solves the problem
- Do not refactor unrelated code in the same change
- Do not add speculative features, abstractions, or error handling for impossible scenarios
- If a test fails, fix the test only if the new behavior is intentionally correct
## Process
- Read existing code before writing new code
- Run tests before and after changes
- Commit messages: imperative mood, concise subject line
| Tool | Target File | Command |
|---|---|---|
| Claude Code | CLAUDE.md | ln -sf AGENTS.md CLAUDE.md |
| Gemini CLI | GEMINI.md | ln -sf AGENTS.md GEMINI.md |
| Cursor | .cursorrules | ln -sf AGENTS.md .cursorrules |
| Copilot | .github/copilot-instructions.md | mkdir -p .github && ln -sf ../AGENTS.md .github/copilot-instructions.md |
Codex reads AGENTS.md natively, so no symlink is needed.
.cursorrules (legacy format) is used intentionally. The newer .cursor/rules/*.mdc format requires YAML frontmatter and is incompatible with plain symlinks.
Run these steps in the project root (the directory containing .git):
If AGENTS.md already exists, show its content to the user and ask whether to overwrite or keep it.
Create AGENTS.md with the template above. If the user wants to customize the content, apply their changes before writing.
For each symlink target in the map above:
<file>.bak) or overwrite.For .github/copilot-instructions.md, run mkdir -p .github first. Use ../AGENTS.md as the symlink source because the file is one directory deeper.
Run ls -la AGENTS.md CLAUDE.md GEMINI.md .cursorrules .github/copilot-instructions.md to show the results to the user.