💡 Summary
A collection of modular agent skills designed to enhance coding workflows for software engineers.
🎯 Target Audience
🤖 AI Roast: “Powerful, but the setup might scare off the impatient.”
Risk: Medium. Review: shell/CLI command execution; outbound network access (SSRF, data egress). Run with least privilege and audit before enabling in production.
dot-agent 🤖
A curated collection of agent skills for coding agents designed to enhance the workflows of Software Engineers (and, let's be honest, me primarily).
Skills provide reusable, domain-specific knowledge and workflows that agents can use when relevant. Works with Cursor, Claude Code, Codex, OpenCode, Antigravity, GitHub Copilot, Amp, Roo Code, and other agent harnesses that support the Agent Skills standard.
What are Skills?
Skills are modular, self-contained packages that extend agent capabilities with:
- Specialized workflows for specific domains
- Tool integrations and instructions
- Domain expertise and best practices
- Bundled resources (scripts, references, assets)
Skills use progressive disclosure - keeping SKILL.md files under 200 lines with detailed content in references/ directories loaded on-demand.
Getting Started 🚀
Install All Skills (Recommended)
Install all skills globally (to ~/.cursor/skills/, ~/.claude/skills/, etc. depending on agent):
npx add-skill siviter-xyz/dot-agent --yes --global --agent cursor
Install Specific Skills
Install individual skills by name:
# Install a single skill npx add-skill siviter-xyz/dot-agent --skill typescript --yes --global # Install multiple specific skills npx add-skill siviter-xyz/dot-agent --skill typescript --skill python --skill astroflare --yes --global
List Available Skills
See all available skills without installing:
npx add-skill siviter-xyz/dot-agent --list
Install to Specific Agents
Target specific agent harnesses:
# Install to Cursor only npx add-skill siviter-xyz/dot-agent -a cursor --yes --global # Install to multiple agents npx add-skill siviter-xyz/dot-agent -a cursor -a claude-code --yes --global
Project-Level Installation
Install skills to your project (committed in repo):
# Install all skills to project npx add-skill siviter-xyz/dot-agent --yes # Install specific skill to project npx add-skill siviter-xyz/dot-agent --skill psi --yes
For more options, see the add-skill documentation.
Sync from Local Repository
When working on dot-agent locally, sync all skills to your local agent installation:
# Sync all skills from local dot-agent repo to Cursor npx add-skill $(pwd) --yes --global --agent cursor # Sync to multiple agents npx add-skill $(pwd) --yes --global --agent cursor --agent claude-code # Sync specific skills only npx add-skill $(pwd) --skill python --skill typescript --yes --global --agent cursor
This is useful when developing or modifying skills locally before publishing.
Available Skills
See skills/README.md for a complete list of available skills.
Project-Specific Configuration
For project-specific rules and commands, commit them in your repository:
- AGENTS.md: Project root or subdirectories for inline rules (works across agent harnesses, PREFERRRED)
- Rules:
.cursor/rules/,.claude/rules/,.codex/rules/, etc. in your repo - Commands:
.cursor/commands/,.claude/commands/,.codex/commands/, etc. in your repo
Keep project-specific content small and focused. Use skills for reusable knowledge.
Progressive Disclosure
All skills follow the progressive disclosure principle:
- SKILL.md < 200 lines - Core principles and when to use
- references/ - Detailed documentation loaded on-demand
- scripts/ - Executable code for deterministic tasks
- assets/ - Files used in output (templates, etc.)
This keeps context windows clean while providing access to specialized capabilities when needed.
Contributing
Skills are designed to be:
- Focused - Each skill covers a specific domain
- Reusable - Work across projects and teams
- Maintainable - Clear structure with progressive disclosure
When contributing:
- Keep SKILL.md under 200 lines
- Use references/ for detailed content
- Follow existing skill structure and the Agent Skills specification
- Prefer agent-harness agnostic skills; if a skill is environment-specific, document this clearly in its frontmatter and/or README
- For scripts inside skills:
- Prefer uv scripts (PEP 723
# /// scriptmetadata with#!/usr/bin/env -S uv run --script) or plain Bash for simple glue - Aim for cross-platform behavior (Linux, macOS; avoid hard-coding shell- or OS-specific paths)
- Recommend installing
uvlocally and running scripts viauv runwhere appropriate
- Prefer uv scripts (PEP 723
- Test installation and usage
References & Inspiration
Pros
- Modular and reusable skills.
- Supports multiple agent harnesses.
- Focus on domain-specific workflows.
Cons
- May require familiarity with multiple agents.
- Limited to the skills provided.
- Dependency on external tools for full functionality.
Related Skills
useful-ai-prompts
A“A treasure trove of prompts, but don’t expect them to write your novel for you.”
fastmcp
A“FastMCP: because who doesn't love a little complexity with their AI?”
Disclaimer: This content is sourced from GitHub open source projects for display and rating purposes only.
Copyright belongs to the original author siviter-xyz.
