AI agent configuration for this repository. For the full architecture reference, see ~/skills/ARCHITECTURE.md.
| Prompt | Purpose | Invokes |
|---|---|---|
/new-project |
Start new project from one-liner idea | @product-designer |
/new-agent |
Create a new agent role | @lead |
/work-on-next |
Pick up next task from plan.md | @lead |
/self-improve |
Improve agent instructions | (meta) |
| Agent | Role |
|---|---|
@product-designer |
Product Designer — drafts spec.md from ideas |
@lead |
Tech lead — orchestrates work, generates plan.md |
@engineer |
Software engineer — dedicated implementation |
@reviewer |
Code reviewer — critical "grill me" feedback |
@refactor |
Architecture & maintainability — small safe refactors |
| Skill | Description |
|---|---|
decompose-task |
Break large tasks into atomic sub-tasks |
diff-check |
Author's cleanup before commit/PR |
tauri-contract |
TS ⇄ Rust contract sync (repo-specific) |
/new-project
> "A lightweight screen recorder like GNOME Screencast"
/work-on-next
Invoke @reviewer with your changes or PR URL.
/self-improve
After making a mistake that should be prevented, ask to update instructions.
npx skills add b1tank/skillscd ~/skills && ./sync.sh ~/opensnippingCreate in .github/skills/<name>/SKILL.md
- Edit files in
~/skills/ - Push:
cd ~/skills && git add . && git commit -m "..." && git push - Re-sync to repos:
./sync.sh <target-repo>
.github/
├── copilot-instructions.md # Repo-wide rules (source of truth)
├── README.md # This file
├── agents/ # Synced from ~/skills + repo-specific
├── prompts/ # Synced from ~/skills + repo-specific
├── skills/ # Repo-specific only
│ └── tauri-contract/
└── archive/ # Old agents (reference)
AGENTS.md (root) # → symlink to .github/copilot-instructions.md
.claude/CLAUDE.md # → symlink to ../.github/copilot-instructions.md
- Golden Architecture Guide — Full reference for agent/skill/prompt design
- b1tank/skills — Shareable skills, agents, prompts