new-mcp-builder
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when…
Use when the user asks to copy, install, refresh, or update a repo-root AGENTS.md file from their canonical global instruction source. Also use when the user says "agent copy" or wants a repo-local AGENTS.md with global operating rules plus project-specific guidance.
$ npx -y skills add shane9coy/Agent-Skill-Architecture-Guide --skill agent-copy --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/agent-copyContext preview
The summary Claude sees to decide when to auto-load this skill.
Use when the user asks to copy, install, refresh, or update a repo-root AGENTS.md file from their canonical global instruction source. Also use when the user says "agent copy" or wants a repo-local AGENTS.md with global operating rules plus project-specific guidance.
name: agent-copy description: Use when the user asks to copy, install, refresh, or update a repo-root AGENTS.md file from their canonical global instruction source. Also use when the user says "agent copy" or wants a repo-local AGENTS.md with global operating rules plus project-specific guidance. metadata: short-description: Copy canonical instructions into a repo and append repo-specific guidance
This skill creates or refreshes a repo-root `AGENTS.md` file by copying the user's canonical global agent instructions, then appending a repo-specific guide for the current project.
Architecture:
canonical global instructions -> repo-local AGENTS.md -> repo-specific tail
Use one `agent-copy` skill for this workflow. Do not create separate runtime-specific variants such as `agent-copy-codex`, `agent-copy-claude`, or `agent-copy-cursor`; change the source path instead.
Default source:
Default target:
Codex is the default because this repo demonstrates a Codex-first setup. If the user's canonical instructions live somewhere else, use that source path instead. Examples include a Claude project instruction file, an OpenCode context file, or another agent-compatible global instruction file.
Only `AGENTS.md` files should be created, copied, or updated unless the user explicitly asks for another instruction file.
The repo-root `./AGENTS.md` file must:
1. Start with an exact copy of the selected canonical global instruction source. 2. Append a repo-specific tail section titled `## Repo-Specific Guide`. 3. Include only repo facts from filesystem inspection or the user's answers. 4. Mark unknowns as `TBD - ask the user before implementing this area.`
Do not paraphrase, summarize, or partially copy the canonical source file.
Check for the default source:
test -f "$HOME/.codex/AGENTS.md" && echo "$HOME/.codex/AGENTS.md"
If the file does not exist and the user wants Codex as the canonical source, tell the user to install the repo's root `AGENTS.md` template into their Codex root first:
mkdir -p "$HOME/.codex" cp AGENTS.md "$HOME/.codex/AGENTS.md"
If the user is not using Codex, ask for the exact path to their equivalent canonical instruction file before copying anything.
Run before editing:
pwd git status --short --branch git branch --show-current git remote -v find . -maxdepth 2 -type f \( -name 'package.json' -o -name 'pnpm-lock.yaml' -o -name 'package-lock.json' -o -name 'yarn.lock' -o -name 'pyproject.toml' -o -name 'requirements.txt' -o -name 'go.mod' -o -name 'Cargo.toml' -o -name '.env.example' -o -name 'README.md' -o -name 'DESIGN.md' \) -print find . -maxdepth 3 \( -name AGENTS.md -o -name CLAUDE.md \) -print
Scan and read existing repo docs/manifests before writing repo-specific details.
If `./AGENTS.md` already exists:
If existing repo instructions conflict with the selected canonical source, keep the canonical source rules unless the user explicitly says the repo is an exception.
Append this section after the global preamble. Populate known fields from the user's project description, `README.md`, `DESIGN.md`, package manifests, lockfiles, env examples, app directory structure, existing scripts, and tests.
Never invent production credentials, live deployment state, database schema, auth policy, or external integrations.
## Repo-Specific Guide ### Project Overview ### Product Goal ### Audience / Users ### Key Directories / Docs / Scripts ### Stack / Architecture ### Local Dev Commands ### Build / Test / Verification ### Branch / Git Workflow ### Boundaries / Do Not Touch ### Implementation Notes
After writing:
SOURCE_AGENTS="${SOURCE_AGENTS:-$HOME/.codex/AGENTS.md}"
head -n "$(wc -l < "$SOURCE_AGENTS")" ./AGENTS.md | cmp "$SOURCE_AGENTS" -
rg -n "Repo-Specific Guide|Project Overview|Product Goal|Stack / Architecture|Build / Test / Verification|Boundaries / Do Not Touch|Implementation Notes" ./AGENTS.md
git status --short --branchIf a non-default source was used, set `SOURCE_AGENTS` to that path before verification.
Report whether the global preamble verification passed.
Final response should include:
Commit only when the user explicitly asked for commit/push or the current task includes publishing.
📄 Get the Full Guide The guide covers agent folder structure, AGENTS.md, SKILL.md anatomy, frontmatter reference, orchestration folders, command-folder equivalents, self-hosted model notes, troubleshooting, and reusable skill patterns.
Repo: shane9coy/Agent-Skill-Architecture-Guide
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when…
Install, create, scaffold, and validate new skills in Codex-first skill folders such as .codex/skills/ and .agents/skills/. Use when adding a skill from a URL,…