How timsonner/agent-skills works
This repository is a catalog of reusable agent skills, operating guides, and a few task-specific scripts—not one integrated runtime. Its clearest end-to-end workflow is the workstation-drive loop: an agent loads the skill, configures a harness if needed, connects to a supplied remote-desktop environment, then interacts through screenshots and UI injection. Other branches cover coding-agent harnesses, platform guidance, administration, security, and PDF-to-Tinker training. The README and sampled files support the relationships shown; many unsampled skills are represented by their documented purpose, without assuming internal wiring.
Main entry points: SKILL.md Workstation Skill · Invoke-MstscConnect.ps1 MSTSC Connector · to_datum.py Datum Builder · mock_tokenizer.py Tokenizer Workaround
Components 21
Workstation Drive
- Workstation Skill
agent-workstation-drive/SKILL.md - Harness Setup
agent-workstation-drive/scripts/Install-ToHarness.ps1 - Linux RDP Skills
gnome-wayland-rdp - Windows UI Inject
windows-ui-inject/SKILL.md - MSTSC Connector
windows-ui-inject/scripts/Invoke-MstscConnect.ps1
Agent Harnesses
- CLI Harness Guides
agy-harness/SKILL.md - Harness Offload
harness-offload/scripts/Invoke-HarnessOffload.ps1 - Copilot SDK
copilot-sdk/SKILL.md - MCP Server Guide
mcp-server/SKILL.md
Platform Skills
- OpenCode Guides
opencode/intro/SKILL.md - Admin Guides
proxmox-system-administration/SKILL.md - Container Guides
wsl-containers/SKILL.md - Desktop Driving
drive-hyprland/SKILL.md - Browser Helpers
read-canvas-browser/SKILL.md
Security Skills
- Pentest Skills
pentest/methodology/SKILL.md
Training Data
- PDF Chunking
pdf-to-tinker-training/scripts/chunk_pdf.py - Tokenizer Workaround
pdf-to-tinker-training/scripts/mock_tokenizer.py - Datum Builder
pdf-to-tinker-training/scripts/to_datum.py - LoRA Trainer
pdf-to-tinker-training/scripts/train_lora.py
Other components
- AI Agent
- Remote Desktop
Connections 19
- AI Agent → Workstation Skill: loads
README.md - Workstation Skill → Harness Setup: sets up
README.md - Workstation Skill → Linux RDP Skills: uses server skill
README.md - Workstation Skill → MSTSC Connector: connects with
README.md - MSTSC Connector → Remote Desktop: opens session
README.md - AI Agent → Windows UI Inject: drives session
README.md - AI Agent → CLI Harness Guides: loads
README.md - AI Agent → Harness Offload: delegates via
README.md - AI Agent → Copilot SDK: integrates with
README.md - AI Agent → MCP Server Guide: builds with
README.md - AI Agent → OpenCode Guides: uses
README.md - AI Agent → Admin Guides: uses
README.md - AI Agent → Container Guides: uses
README.md - AI Agent → Desktop Driving: uses
README.md - AI Agent → Browser Helpers: uses
README.md - AI Agent → Pentest Skills: uses
README.md - PDF Chunking → Datum Builder: chunks feed
pdf-to-tinker-training/scripts/to_datum.py - Datum Builder → Tokenizer Workaround: imports workaround
pdf-to-tinker-training/scripts/to_datum.py - LoRA Trainer → Tokenizer Workaround: imports workaround
pdf-to-tinker-training/scripts/train_lora.py