Claude Code

User-Level Skills

Reusable how-to guides that extend Claude's capabilities — available in every session, every project

Location: ~/.claude/skills/

ai-servicers.com

What Are Skills?

A skill is a SKILL.md file that Claude reads when a relevant task comes up.

Think of them as onboarding guides. Instead of Claude having to figure out how things work from scratch each session, a skill hands it the procedures, naming standards, gotchas, and tool integrations for a specific domain.

Zero boot cost. Skill metadata (name + description) is always loaded. The full SKILL.md body is only read when the skill is triggered — keeping context lean.

How Skills Get Triggered

1
Metadata Loaded
At session start, Claude reads the name and description of every skill. This costs ~100 words per skill — not the full content.
2
Trigger Matched
User says "deploy", "install", "create a presentation", "troubleshoot" — the description tells Claude which skill to invoke.
3
SKILL.md Read
Claude reads the full skill body — workflow steps, commands, gotchas, examples. Now it has expert-level guidance for the task.
4
Resources Used
The skill can reference bundled scripts, templates, or reference docs. Claude loads only what it needs for the current task.

Result: 0 tokens at boot • full depth on demand • context stays focused

Category

Infrastructure & DevOps Skills

Infrastructure
naming-validator
Validates resource names against standards BEFORE creating anything. Enforces lowercase-with-hyphens for services, snake_case for databases. Triggers on "install", "create", "deploy".
Infrastructure
secrets-security
Secrets management and security hardening. Ensures credentials go to $HOME/secrets/ not project dirs. Covers Bitwarden, credential rotation, security audits.
Infrastructure
gitlab-integration
GitLab repository management, issues, merge requests, and CI/CD pipelines for gitlab.ai-servicers.com. Includes pipeline safety rules to prevent concurrent run pile-ups.
Infrastructure
gitlab-board
GitLab issue board management using checkbox-based workflows. Supports decision, info, and feature issues with spec-kit methodology for AI/human collaboration via card status.
Infrastructure
logging-observability
Observability stack: Loki log aggregation, Promtail collection, Grafana dashboards, Netdata system metrics, Dozzle container log viewer. Used when deploying or troubleshooting services.
Category

AI Workflow Skills

AI Workflow
project-context → details
Persistent project knowledge in docs/context/ (7 files). Auto-triggered via Stop/SubagentStop hooks with smart git diff. Ensures agents maintain architecture, decisions, security, operations, testing, gotchas, and specs across sessions.
AI Workflow
createsolution
Multi-AI solution design: Gemini + Codex + Claude generate perspectives in parallel, synthesized into a final solution with executive web presentation deployed to nginx.
AI Workflow
createplan
Multi-AI project planning: all three AIs generate plans, Claude synthesizes a draft, peer review from Gemini & Codex, final plan with presentation. Produces agent assignments and phase breakdown.
AI Workflow
createreview
AI Board Review for critical analysis: brainstorm approaches, compare against best practices, focus on simplicity. Use to evaluate technology choices or architecture decisions.
AI Workflow
executeplan
Bridges planning to execution: creates GitLab project + Kanban board from createsolution/createplan artifacts. Generates trackable cards so AI and human collaborate via board status.
AI Workflow
agent-collaboration
Guide for delegating to other agents. Defines which agents handle what (claude-administrator for infra, gemini for fast lookups, Review Board for analysis). Max 1 hop rule enforced.
Category

AI Workflow Skills (continued)

AI Workflow
alignment-review
Investigates AI agent misalignment caused by context files. Finds the directive that caused unexpected behavior, proposes a fix, and creates a GitLab card for human approval.
AI Workflow
ai-communication
AI-to-AI communication: file sharing via MinIO aichat-files bucket and real-time dispatch to Gemini/Codex/Claude via MCP Review Board. Covers async and sync communication patterns.
AI Workflow
cchat
Inter-CLI Matrix chat via MCP tools (chat_send, chat_read, chat_who). Routes through aiagentchat-gateway. Used for communication between Claude Code CLI sessions and AI agents.
Category

Development & Specialist Skills

Development
python-pro
Senior Python 3.11+ patterns: type hints, async/await, pytest, dataclasses, mypy. Production-ready code with complete type coverage. Triggers for Python development tasks.
Development
typescript-pro
Advanced TypeScript: generics, conditional types, mapped types, tRPC, tsconfig, discriminated unions. Full-stack type safety patterns and production-grade TS development.
Development
postgres-pro
Expert PostgreSQL: EXPLAIN ANALYZE, JSONB operations, streaming replication, VACUUM tuning, PostGIS, pgvector. For advanced query optimization and DBA tasks.
Development
frontend-design-pro
Production frontend with bold aesthetics (glassmorphism, brutalism, neumorphism, etc.) and real images via Unsplash/Pexels direct links or custom AI image prompts. No AI slop.
Development
architecture-diagram-creator
Comprehensive HTML architecture diagrams: SVG data flows, business objectives, features, technical architecture, and deployment docs. Triggers on "create architecture diagram" requests.
Development
feature-forge
Requirements specialist for defining features. Structured workshops using PM + Dev perspectives, EARS format specifications, user stories, and acceptance criteria.
Development
spec-miner
Reverse-engineering specialist: extracts specifications from existing codebases, creates documentation for legacy or undocumented systems. Arch Hat + QA Hat perspectives.
Category

Meta & Utility Skills

Meta
skill-creator
Guide for creating new skills. Covers anatomy, progressive disclosure, trigger writing, bundled resources, and packaging workflow. Triggers on "create a skill" or "update skill".
Meta
revealjs-presentation
Professional reveal.js HTML presentations with strong visual design, varied layouts, and consistent theming. Triggers on "slides", "presentation", "deck", "reveal.js", HTML slideshow.
Utility
openmemory
Persistent memory across sessions using mem0 with Gemini embeddings. Stores user preferences, technical decisions, patterns, and lessons learned. Retrieves via semantic search.
5 Infrastructure
9 AI Workflows
7 Development
3 Meta / Utility
24 Total

How to Create a New Skill

Anatomy of a Skill

  • Directory: ~/.claude/skills/my-skill-name/
  • Required file: SKILL.md — frontmatter with name + description triggers invocation
  • Optional: references/ for docs, scripts/ for automation, assets/ for templates

SKILL.md Frontmatter (required)

  • name: — machine name, matches directory name
  • description: — plain English trigger guide. Claude reads this to decide when to invoke. Include keywords.

Progressive Disclosure Pattern

  • Metadata (name + description) — always loaded, ~100 tokens per skill
  • SKILL.md body — loaded when triggered, under 5k words
  • Bundled resources — loaded as needed by Claude during task execution

Use the skill-creator skill for the full guided workflow.

User-Level Skills

24 skills in ~/.claude/skills/ — available in every session, every project.

Zero tokens at boot. Metadata is indexed at startup. Full skill content loads only when triggered. Context stays focused — depth appears exactly when needed.

Skills cover infrastructure enforcement, AI workflow orchestration, specialist development knowledge, and meta-tools for extending the system itself.

24 Total Skills
0 Boot Tokens
4 Categories

ai-servicers.com