ai-coding-tooling

Shared agent configuration and skills for AI coding tools (Claude Code, Cursor, Windsurf, Gemini CLI, etc.).

Installation

Prerequisites

• Unix-like shell (macOS/Linux)
• make installed

1. Clone the repository

git clone <repo-url>
cd ai-coding-tooling

2. Create the symlinks

make link

Existing files or symlinks are skipped — nothing is overwritten.

3. Set up global agent configuration

Open this project in Claude Code and run:

/agent-setup

This will symlink AGENTS.global.md to the agent's global config file and install all global skills.

4. Install extended skill files

After the global skills are installed (step 3), link the extended skill files into the installed skill directories:

make link-extended

5. Install the status line script

Run this separately (not included in make link) to install the Claude Code status line script to ~/.claude/statusline-command.sh. It displays the active model, effort level, current directory, git branch, context remaining, and token usage:

[Claude Sonnet 4.6 - **High**] 📁 my-project git:(main) | 82% | 4321 tokens

The script is copied from config/statusline-command.sh. If the destination already exists it is skipped:

make install-statusline          # skip if file already exists
make install-statusline FORCE=1  # overwrite with the version from this repo

To customize the status line without losing your changes on the next FORCE=1 run, edit ~/.claude/statusline-command.sh directly — omit the FORCE=1 flag and it will never be overwritten.

6. Remove symlinks

make unlink

Only removes symlinks — real files are never deleted.

Skills

Skills are reusable agent instructions that extend AI coding tools with specialized workflows. They are grouped below by source.

Project-Local Skills (.agents/skills/)

These skills live in .agents/skills/ and are auto-loaded when the project is opened in Claude Code. No installation step required.

Skill	Description
agent-setup	Sets up global agent configuration by symlinking AGENTS.global.md and installing all global skills. Multi-agent — supports Claude Code, Gemini CLI, and others via agent reference files.
generate-knowledge-base	Reads files or folders, extracts intelligence, and produces a comprehensive Markdown knowledge note saved to the Obsidian vault.
raindrop-kb-convert	Converts a Raindrop.io bookmark collection into a consolidated knowledge base in the Obsidian vault. Clusters bookmarks by topic and generates deduplicated .md files per cluster.
skill-architect	Expert guide for designing and building high-quality skills through structured conversation. Source: Tech Leads Club. Installed locally.
skill-manager	Installs or updates skills in an agent's global skills directory. Handles fresh installation (Tech Leads Club via npx or local project via symlink) and updates externally installed skills by reinstalling from their vendor registry and re-applying extended skill symlinks. Prompts for install vs update if intent is ambiguous.

Source: This Project (ai-coding-tooling)

Maintained here and installed globally via make link / /agent-setup. These are the only skills you should modify:

Skill	Description
architecture-evaluate ⭐	Creates or updates the three mandatory project context files (PROJECT_DETAILS.md, ARCHITECTURE.md, PIPELINE.md) inside docs/. Also supports package mode for evaluating individual packages/modules, generating a scoped CLAUDE.md inside the package directory. Run this when onboarding a new project or when context files are missing.
code	Apply coding guidelines when writing code. Thin delegator to coding-guidelines (TLC). Pairs with code-review for the /code + /code-review naming pattern.
code-review	Performs comprehensive code reviews covering architecture, performance, code quality, API design, and security. Reviews local workspace changes by default, or a GitHub PR when a PR number is provided. Also performs standalone Performance Audits (full-codebase P0–P3 findings report) when triggered by performance audit phrases.
documentation-upsert	Updates all project documentation by inspecting the git workspace for modified files. Detects new packages and triggers scoped architecture evaluation. Updates inline API docs, root context files, and base docs/ files.
tech-debt-report	Documents tech debts in docs/tech-debts/ and maintains an anti-pattern index in docs/TECH_DEBTS.md so agents avoid replicating bad patterns.
tech-reference-add ⭐	Adds technology-specific reference files across all skills and extends qualifying global skills. Run this when adding a new framework or language to a project's stack.
tests	Writes and maintains tests — unit, integration, and coverage analysis.
tests-code-review	Reviews test code quality, coverage patterns, and maintainability. Supports local workspace and GitHub PR review modes.

⭐ Highlighted skills:

• architecture-evaluate — The recommended first step for any new or onboarded project. It generates the context files in docs/ that agents load progressively based on task relevance.
• tech-reference-add — The recommended way to extend the tooling for a new technology. It propagates tech-specific reference files into all relevant skills (code review, coding guidelines, tests, etc.) in one step.
• tech-debt-report — Documents known anti-patterns so agents avoid replicating them. The docs/TECH_DEBTS.md index is loaded automatically when writing or reviewing code.

Skill Aliases

Aliases are thin delegators that provide shorter slash commands for existing skills. The original skill remains unchanged and fully functional.

Alias	Original Skill	Description
code	coding-guidelines	Short command for /coding-guidelines. Pairs with /code-review.

Personal Skills (personal/)

You can add private, local-only skills that are never committed to git. Create a personal/ directory at the project root and add skill subdirectories inside it — each must contain a SKILL.md file following the same structure as skills/.

personal/
  my-private-skill/
    SKILL.md

/agent-setup and make link both auto-discover and install everything in personal/ via symlink. Personal skills installed via make link are removed with make unlink. These skills are labeled Personal (local-only) in the /agent-setup install summary and are never listed in AGENTS.md.

The personal/ directory is gitignored — nothing inside it is tracked or committed.

Recommended MCP Servers

MCP Server	Purpose	Used by
Context7	Fetches up-to-date documentation and code examples for any library. Provides authoritative raw material when generating technology-specific reference files.	tech-reference-add (Step 7)

Context7 is optional but strongly recommended. When available, tech-reference-add queries it for official documentation to ground reference files in current best practices rather than relying solely on LLM training data. If unavailable, the skill falls back to the agent's own knowledge.

Source: Tech Leads Club

Installed globally by /agent-setup. Treated as read-only — do not edit these directly:

Skill	Description
codenavi	Pathfinder for navigating unknown codebases. Investigates with precision, implements surgically, and never assumes. Use when fixing bugs, implementing features, refactoring, or investigating flows in unfamiliar territory.
coding-guidelines	Behavioral guidelines to reduce common LLM coding mistakes. Applied when writing or reviewing code.
docs-writer	Writing, reviewing, and editing documentation and .md files.
security-best-practices	Language and framework specific security reviews (Python, JavaScript/TypeScript, Go).
subagent-creator	Guide for creating AI subagents with isolated context for complex multi-step workflows.
technical-design-doc-creator	Creates comprehensive Technical Design Documents (TDD) following industry standards.
web-design-guidelines	Reviews UI code for accessibility, design, and best-practices compliance.