Skip to content
FSS Knowledge Base

AI Skills Ecosystem

Section titled “AI Skills Ecosystem”

Skills are version-controlled, auditable, testable instruction packages that define how AI should perform specific tasks. They represent a fundamental shift from one-off prompting to reusable, composable AI capabilities.

Why Skills Matter

Section titled “Why Skills Matter”
  • As significant as spreadsheets for productivity enhancement — accessible to non-programmers
  • No coding needed, no APIs needed — just clear instructions in plain text
  • Version-controlled, auditable, testable, and attributable
  • Simple structure creates a compounding flywheel of knowledge and capabilities
  • Published as open standard at agentskills.io
  • MCP defines how Claude calls external systems; Skills define what Claude should do

Skill Structure

Section titled “Skill Structure”

A skill is a folder containing:

.claude/skills/skill-name/
├── SKILL.md          # Core instructions (required)
├── scripts/          # Helper scripts (optional)
├── templates/        # Templates (optional)
└── resources/        # Reference docs (optional)

SKILL.md Best Practices

Section titled “SKILL.md Best Practices”
  • Naming: use gerund form (verb + -ing): processing-pdfs, reviewing-code
  • Description: state what it does AND when to use it
  • Progressive disclosure: only load heavy context when necessary
  • !command syntax injects shell command output into skill content
  • context: fork in frontmatter runs skill in isolated subagent
  • PreToolUse, PostToolUse, Stop hooks can be defined in frontmatter

Invocation

Section titled “Invocation”
  • /skill-name from Claude Code CLI
  • Skills auto-reload when SKILL.md changes
  • Simple commands stay in .claude/commands/ as single .md files

Meta-Tools for Building Skills

Section titled “Meta-Tools for Building Skills”

Skill-Creator (Anthropic Official)

Section titled “Skill-Creator (Anthropic Official)”

Skill-Updater

Section titled “Skill-Updater”

Skills Factory (CCGG Community)

Section titled “Skills Factory (CCGG Community)”
  • Generates complete skills from brain dumps in minutes
  • Handles research, architecture, and deployment
  • https://ccggskool.com

Skilz Universal Installer

Section titled “Skilz Universal Installer”

Skills Repos and Directories

Section titled “Skills Repos and Directories”
SourceURL
Anthropic Officialhttps://github.com/anthropics/skills/tree/main
Awesome Claude Skills (Composio)https://github.com/ComposioHQ/awesome-claude-skills
Awesome Claude Skills (VoltAgent)https://github.com/VoltAgent/awesome-claude-skills
Awesome Claude AIhttps://awesomeclaude.ai/
Jimmy Song’s Directoryhttps://jimmysong.io/ai/awesome-claude-skills/
SkillsMP (87K+ skills)https://skillsmp.com/docs/api
Skillz Marketplacehttps://skillzwave.ai/

Priority Skills to Install/Build

Section titled “Priority Skills to Install/Build”

For Development

Section titled “For Development”

Superpowers (INSTALL FIRST - Highest Priority)

Section titled “Superpowers (INSTALL FIRST - Highest Priority)”

The most mature and widely adopted CC skills framework

Stats: 47.7K GitHub stars, official Anthropic marketplace acceptance (Jan 2026)

What it provides:

  • 14 composable skills that activate based on context
  • Three-phase workflow: brainstorm → plan → execute
  • Enforces TDD (write failing test → implement → verify pass)
  • Two-stage review: spec compliance + code quality
  • Fresh subagent per task with full context
  • Prevents Claude from cutting corners or premature “done”

The 14 Core Skills:

  1. brainstorming — Socratic requirements refinement
  2. writing-plans — Granular task decomposition with TDD
  3. executing-plans — Step-by-step execution with checkpoints
  4. subagent-driven-development — Fresh subagent + two-stage review
  5. dispatching-parallel-agents — Concurrent execution
  6. test-driven-development — RED-GREEN-REFACTOR enforcement
  7. systematic-debugging — 4-phase root cause analysis
  8. verification-before-completion — Evidence-based done checks
  9. using-git-worktrees — Isolated branches
  10. requesting-code-review — Review against spec
  11. receiving-code-review — Handle feedback
  12. finishing-a-development-branch — Merge and cleanup
  13. using-superpowers — Meta-skill
  14. writing-skills — Author new custom skills

Installation:

Terminal window
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

Usage:

Terminal window
/superpowers:brainstorm    # Socratic Q&A, explore approaches
/superpowers:write-plan    # Generate granular TASKS.md
/superpowers:execute-plan  # Run with TDD + subagent review

When to use:

  • Any task > 30 minutes
  • New features requiring planning
  • Complex refactors
  • When quality and verification matter

When to skip:

  • Quick fixes (< 5 min)
  • Markdown KB work
  • Trivial tasks
  • Overhead (10-20 min planning) outweighs value

Resources:

Other High-Priority Skills

Section titled “Other High-Priority Skills”
  • frontend-design (Anthropic): modern design principles, prevents “AI Slop” UI
  • Playwright Skill: E2E testing automation — https://github.com/lackeyjb/playwright-skill
  • ffuf-web-fuzzing: AI security auditor
  • taskfix: custom skill for task-list-driven fix-verify workflow (build this)

For SMART DEBT Mission

Section titled “For SMART DEBT Mission”
  • SvelteKit component generators and refactoring
  • Debt calculator APIs and financial education prototypes
  • Content generation and branding
  • Smart debt coach automated service

For Workflow

Section titled “For Workflow”
  • verify-deploy: smoke test runner before marking tasks complete (build this)
  • codebase-audit: proactive health check (stale refs, config drift, TODOs)

Key Articles

Section titled “Key Articles”