Agent Skills Repository for AI Assistants

A modular, production-grade collection of 2-Tier Progressive Disclosure AI Agent Skills designed for LLM assistants (Claude, ChatGPT/OpenAI, and Google Gemini).

---

📂 Skill Categories

This repository is organized into domain categories. Each skill contains a Tier 1 SKILL.md prompt module and on-demand Tier 2 technical references in references/.

🏋️ Athletic Performance & Sports Coaching (coach/)

Skill Name	Directory	Primary Focus & Domain	Key Metrics & Methods
Cycling Coach	coach/cycling/	Power-based cycling, workout analysis, race pacing, event optimization	Coggan Z1-Z7, FTP, NP, IF, TSS, TSB, Aerobic Decoupling, Intervals.icu
Running Coach	coach/running/	VDOT pacing, marathon periodization, impact stress management	Jack Daniels VDOT, E/M/T/I/R paces, GAP, Cadence (spm), rTSS, Pace Decoupling
Swimming Coach	coach/swimming/	CSS threshold pacing, SWOLF efficiency, stroke mechanics, shoulder health	Critical Swim Speed (CSS), SWOLF score ($\text{Time} + \text{Strokes}$), DPS, sTSS
Rowing Coach	coach/rowing/	2K/5K split pacing, ergometer & water technique, drag factor, spine & rib health	Split /500m, Stroke Rate (spm), Wattage, Drag Factor, Drive:Recovery (1:2), rTSS
Weight Training Coach	coach/weight-training/	RIR/RPE autoregulation, mesocycle design, powerlifting, concurrent integration	Reps in Reserve (RIR 0-4), 1RM %, Volume Load, VBT, mTOR/AMPK balancing
Triathlon Coach	coach/triathlon/	Multi-sport load balancing, brick workouts, transition mechanics (T1/T2)	Composite Load ($\text{cTSS}$), T2 Brick runs, T1/T2 setup, Sprint/Olympic pacing
Ironman Coach	coach/ironman/	70.3 & 140.6 ultra-endurance pacing, high-carb gut training, 21-day taper	Bike IF caps ($\le 0.73$), 80-120g/hr carbs, 800-1200mg/hr sodium, Special Needs
XC Skiing Coach	coach/xc-skiing/	Skate & Classic subtechniques, roller skiing, cold-air airway safety	V1/V2/V2-Alt shifting, Double Poling core flex, Roller-skiing, Vasaloppet
Sports Nutrition Coach	coach/sports-nutrition/	Evidence-based sports nutrition, macro periodization, intra-workout fueling, RED-S	Carbohydrate scaling (3-12g/kg), 30-120g/h carbs, 300-1000mg/h sodium, EA $\ge 45$, AIS supplements

🛡️ System Administration & Security (sysadmin/)

Skill Name	Directory	Primary Focus & Domain	Key Metrics & Methods
Firewall Admin	sysadmin/firewall-admin/	Enterprise firewall rule design, stateful packet filtering, NAT, VPN tunnels, threat inspection	Zone Security (DMZ/LAN/WAN), SNAT/DNAT, IKEv2 IPsec, WireGuard, nftables, IPS/DPI
Proxmox Admin	sysadmin/proxmox-admin/	Proxmox VE 9.x node/cluster lifecycle, ZFS/Ceph Squid storage, SDN Fabrics, VM/LXC orchestration	Corosync 3, pvecm, qm/pct, ZFS OpenZFS 2.3, Ceph 19.2, PBS, HA Affinity

(More categories—such as Software Engineering, Writing & Communications, and Productivity—can be added as dedicated skill subdirectories!)

---

🔌 Live Athletic Data Integration (MCP Server)

To enable your AI coaching skills to directly fetch live workouts, fitness metrics (CTL, ATL, TSB), wellness signals (HRV, RHR, Sleep), and activity stream data from Intervals.icu, pair these skills with the intervals.icu-mcp server:

• Repository: https://github.com/jelmervdm/intervals.icu-mcp
• Capability: Provides live tools and resources for AI agents to query Intervals.icu API endpoints (fitness trends, power/HR stream data, workout creation, and event calendar sync).
• Supported Environments: Claude Desktop, Antigravity IDE, Cursor, and VS Code.

---

🛠️ Repository Architecture

All skills strictly conform to the Agent Skills Specification (2-Tier Progressive Disclosure):

• Tier 1 (SKILL.md): Contains YAML frontmatter, core prompt triggers, 5-step operational workflow, coaching/domain philosophy, routing table, and non-negotiable safety guardrails.
• Tier 2 (references/*.md): Deep technical reference files loaded on-demand when specific sub-topics are triggered.

---

🚀 Installation & Integration Options

Tip

💡 Zero-Dependency Native Usage (No CLI or Plugins Required)

Every skill in this repository is simply a plain Markdown file (SKILL.md). You do not need Node.js, npx, plugins, or marketplace tools to use them. You can simply copy the contents of any SKILL.md into your AI assistant's System Prompt, Project Context, or Custom GPT instructions!

1. Manual File Copy / Drop-In (Simplest & Zero Setup)

• VS Code & Antigravity IDE: Create a folder named .agents/skills/ in your project root and copy skill folders into it. The IDE agent automatically reads it.
• Claude Projects / Desktop: Open Claude.ai, create a Project, and upload SKILL.md and references/*.md directly as project context.
• ChatGPT / Custom GPTs: Paste SKILL.md into Custom GPT System Instructions and upload references/ into Knowledge Files.
• Local User Directory: cp -r ./coach/* ~/.claude/skills/ or ~/.agents/skills/.

2. Optional CLI & GitHub Install Shortcuts

• Claude Code CLI:

  claude plugin install https://github.com/jelmervdm/skills

• Agent Skills CLI (Installs to 40+ AI Agents including Cursor, Windsurf, Claude):

  npx agent-skills-cli@latest add @jelmervdm/skills

• skills.sh Ecosystem:

  npx skills add jelmervdm/skills

• Global User Directory (Local Copy):

  mkdir -p ~/.claude/skills
  cp -r ./coach/* ~/.claude/skills/

2. VS Code & Antigravity IDE

• Workspace Skill Discovery: Copy desired skill folders into .agents/skills/ within your project root (e.g. .agents/skills/cycling/ or .agents/skills/coach/cycling/). Agentic IDEs automatically detect and auto-load these skills.
• Direct Workspace Referencing: Tag skill files directly in agent prompts using @coach/<sport>/SKILL.md.
• Gemini API: Load SKILL.md content directly into GenerativeModel(system_instruction=...).

3. Claude Desktop & Projects

• Claude Projects: Create a project, upload the desired SKILL.md and references/*.md files, and set project instructions.
• Claude Desktop: Paste SKILL.md into system prompt or attach files directly into conversation context.

4. OpenAI (ChatGPT / Custom GPTs / API)

• Custom GPTs: Paste SKILL.md instructions into the GPT Configuration box, and upload references/*.md into the GPT Knowledge Files.
• OpenAI API: Load SKILL.md as the system message string.

---

🤝 Community & Governance

We welcome contributions across all skill domains! To maintain high repository standards and foster a welcoming community:

• 📋 Contributing Guidelines: Rules for adding skills, progressive disclosure standards, and validation commands.
• 🛡️ Code of Conduct: Community standards adapted from the Contributor Covenant v2.1.
• 🔒 Security Policy: Disclosure policy for reporting security issues and prompt injection vectors.
• 💡 Propose a New Skill: Submit proposals for new specialized skills.
• 🐛 Report a Bug: Report skill errors, broken links, or validator issues.

---

⚖️ License & Credit

This repository is licensed under the MIT License.
Created and maintained by Jelmer (@jelmervdm).
Feel free to fork, adapt, and expand into other domains! Please retain copyright credit in forks as required by the MIT License.