Skip to content

naijoug/skills

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Skills Collection

Personal collection of AI coding skills. Skills are classified by directory under skills/.

Project Structure

.
├── scripts/
│   ├── skills-linker              # Install/uninstall skills (CLI + fzf TUI)
│   └── tui                        # Shortcut for `skills-linker tui`
└── skills/
    ├── global/
    │   └── in-english/
    ├── cron/
    │   ├── daily-til/
    │   ├── daily-trending/
    │   └── weekly-retro/
    ├── auto/
    │   └── skill-smith/
    └── manual/
        ├── plan/
        │   ├── teaching/
        │   ├── code-reading/
        │   └── test-case/
        ├── review/
        │   ├── pr/
        │   ├── api-design/
        │   └── refactor/
        ├── research/
        │   └── web-search/
        ├── growth/
        │   ├── personal-coach/
        │   ├── engineering/
        │   └── debugging-kata/
        ├── tooling/
        │   ├── openclaw/
        │   └── who-am-i/
        └── meta/
            └── example/

Skill Categories

Category comes from the directory path, not skill.yaml:

Category Meaning Typical examples
skills/global/<skill> Install globally and auto-inject into agent instructions in-english
skills/cron/<skill> Trigger from recurring automation or heartbeat jobs, not manual invocation daily-til, daily-trending, weekly-retro
skills/auto/<skill> Auto-trigger helper that activates after installation skill-smith
skills/manual/<group>/<skill> Only runs when explicitly selected or clearly requested most task skills

Behavior:

  • global and auto skills are auto-injected by skills-linker when they provide inject.md
  • cron skills are installed normally but are intentionally not auto-injected; they are driven by scheduled triggers
  • manual skills are the ones exposed by ./scripts/ng
  • manual subgroup comes from the second directory level under skills/manual/

Recommended manual subdirectories:

  • plan for planning, mapping, and decomposition skills
  • review for review, critique, and safe-change skills
  • research for search, trends, and information synthesis
  • growth for practice, coaching, and retrospectives
  • tooling for local tools and service operations
  • meta for skill-authoring or self-profile skills

Installation

Supported Tools

Tool Global Project
Codex ~/.codex/skills {project}/.codex/skills
ChatGPT (alias to Codex) ~/.codex/skills {project}/.codex/skills
Claude Code ~/.claude/skills {project}/.claude/skills
Amp ~/.agents/skills {project}/.agents/skills

Interactive TUI (recommended)

Requires fzf.

./scripts/tui
# Or allow auto-install via Homebrew when fzf is missing
./scripts/tui --auto-install-fzf

TUI flow:

  1. Enter a unified control panel (persistent, does not exit after one action)
  2. Adjust: action, target, policy, skills
  3. Select execute now to run
  4. After execution, stay in the panel for further operations
  5. Select quit to exit

Skills selection supports right-side SKILL.md preview. Install defaults to selecting all skills.

CLI

# List supported tools
./scripts/skills-linker tools

# List available skills
./scripts/skills-linker list

# List only manual skills
./scripts/skills-linker list --category manual

# Install to Claude Code global skills
./scripts/skills-linker install --tool claude --scope global \
  pr personal-coach

# Install to project-level
./scripts/skills-linker install --tool codex --scope project --project-root .

# Uninstall
./scripts/skills-linker uninstall --tool claude --scope global pr

# Check status
./scripts/skills-linker status --tool claude --scope global

Options:

  • --tool NAME — codex | chatgpt | claude | amp | custom (default: codex)
  • --scope SCOPE — global | project (default: global)
  • --mode MODE — symlink | copy (default: symlink)
  • --auto-install-fzf — allow tui to install fzf via Homebrew when missing
  • --force — replace conflicting targets
  • --with-slash-commands — also install slash command wrappers for manual skills
  • --json — JSON output

Slash Commands

--with-slash-commands writes one wrapper file per manual skill into the agent's slash command directory, so you can trigger the skill by typing /<skill-name> in the chat:

Tool Global Project
Claude Code ~/.claude/commands/<skill>.md {project}/.claude/commands/<skill>.md
Codex / ChatGPT ~/.codex/prompts/<skill>.md {project}/.codex/prompts/<skill>.md

Each wrapper file contains a marker comment so the linker can refresh or remove only the files it created — pre-existing files with the same name are left alone unless --force is passed. Uninstall always cleans up matching managed wrappers.

Notes:

  • global skills are best installed in global scope so every agent can see them.
  • global and auto categories may inject extra instructions into AGENTS.md / CLAUDE.md.
  • global scope is usually more convenient because you don't need to reinstall in each project.

Manual Entry Point

manual skills now have a dedicated helper. It groups skills by their directory under skills/manual/:

# Show the category plan
./scripts/ng plan

# List manual skills
./scripts/ng list

# Pick a manual skill interactively (fzf if installed)
./scripts/ng

# Show how to trigger one specific skill
./scripts/ng pr

Creating a New Skill

mkdir -p skills/my-skill
mkdir -p skills/manual/plan/my-skill
cat > skills/manual/plan/my-skill/SKILL.md << 'EOF'
---
name: my-skill
description: Use when [specific trigger conditions]
---

# My Skill

## Overview
What this skill is for.

## When to Use
- Trigger 1
- Trigger 2

## Steps
1. Do X
2. Do Y
EOF

Suggested skills/manual/plan/my-skill/skill.yaml:

id: my-skill
version: 1.0.0
title: My Skill
summary: Use when [specific trigger conditions]
kind: prompt_only

Trigger Evaluation

This repo includes a trigger evaluation workflow for testing skill trigger recall/precision.

  • Per-skill examples: skills/**/references/trigger-examples.md
  • Export + scoring: scripts/trigger_examples_tool.py
  • Runner: scripts/run_trigger_eval.sh
  • HTML report: scripts/trigger_eval_report.py
# Smoke test (perfect predictor)
./scripts/run_trigger_eval.sh --mode perfect --no-details

# Include non-manual skills in dataset when needed
python3 ./scripts/trigger_examples_tool.py --include-non-manual summary

# Custom predictor
./scripts/run_trigger_eval.sh --mode custom \
  --predict-cmd 'python3 "$ROOT_DIR/scripts/predictor_adapter_template.py" --input "$CASES_FILE" --output "$PREDS_FILE"'

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors