Plaited Skills Installer

Install Plaited skills for AI coding agents supporting the agent-skills-spec.

Requirements

• macOS / Linux: Works out of the box
• Windows: Requires one of:
  • WSL (Windows Subsystem for Linux) - recommended
  • Developer Mode enabled (for symlink support)

The installer uses symlinks to avoid duplicating skill files across multiple agents. Windows requires special permissions or WSL for symlink creation.

Installation

For agents supporting the agent-skills-spec (Gemini CLI, GitHub Copilot, Cursor, OpenCode, Amp, Goose, Factory, Codex, Windsurf, Claude Code):

# Install for multiple agents
curl -fsSL https://raw.githubusercontent.com/plaited/skills-installer/main/install.sh | bash -s -- --agents claude,gemini

# Install a specific project
curl -fsSL https://raw.githubusercontent.com/plaited/skills-installer/main/install.sh | bash -s -- --agents claude --project development-skills

Or clone and run locally:

git clone https://github.com/plaited/skills-installer.git
cd skills-installer
./install.sh                              # Interactive mode (multi-select agents)
./install.sh --agents claude,gemini       # Install for multiple agents
./install.sh --agents cursor --project agent-eval-harness  # Specific project
./install.sh --list                       # List available projects
./install.sh --uninstall                  # Remove all

Architecture

Skills are installed to a central .plaited/skills/ directory and symlinked to each agent's skills directory:

.plaited/skills/                          # Central storage (single copy)
  skill-name@org_project/

.claude/skills/                           # Symlinks
  skill-name@org_project -> ../../.plaited/skills/skill-name@org_project

.gemini/skills/                           # Symlinks
  skill-name@org_project -> ../../.plaited/skills/skill-name@org_project

This approach:

• Saves disk space when supporting multiple agents
• Ensures all agents use identical skill versions
• Makes updates simpler (update once, all agents see changes)

Note: Modifying files through symlinks affects all agents sharing that skill.

Supported agents:

Agent	Skills Directory
gemini	.gemini/skills/
copilot	.github/skills/
cursor	.cursor/skills/
opencode	.opencode/skill/
amp	.amp/skills/
goose	.goose/skills/
factory	.factory/skills/
codex	.codex/skills/
windsurf	.windsurf/skills/
claude	.claude/skills/

Available Projects

Project	Source
development-skills	plaited/development-skills
agent-eval-harness	plaited/agent-eval-harness
plaited	plaited/plaited

Skill Scoping

Skills are automatically scoped during installation to prevent naming collisions when multiple projects provide skills with the same name.

Skill folders are renamed using the pattern: <skill-name>@<org>_<project>

Example: typescript-lsp from plaited/development-skills becomes typescript-lsp@plaited_development-skills

Inherited skills (already scoped from another project) preserve their original scope to prevent double-scoping.

Note: Org and project names may contain alphanumeric characters, dots, hyphens, and underscores. Some tools using dot-notation for namespacing may interpret dots specially.

Replace-on-Install Behavior

Running ./install.sh replaces existing skill folders with fresh copies from the source repository. This ensures skills are always up-to-date but will overwrite any local modifications. The installer will display "Replaced" for skills that were updated and "Installed" for new skills.

Note: If you have local modifications you want to preserve, back them up before reinstalling.

Project Details

development-skills

Development tools including:

• typescript-lsp - LSP-based code exploration for TypeScript/JavaScript
• code-documentation - Code documentation generation
• validate-skill - Skill validation tools

Source: plaited/development-skills

agent-eval-harness

CLI tool for evaluating AI agents by capturing execution trajectories:

• Capture full execution traces (thoughts, messages, tool calls, plans)
• Run pass@k evaluations with customizable grading functions
• Schema-driven adapters for any CLI agent producing JSON output
• Unix-style composable pipelines (run → extract → grade → format)

Source: plaited/agent-eval-harness

plaited

Development tools for the Plaited behavioral programming framework:

• loom - Templating and component library
• plaited-behavioral-core - Behavioral programming patterns
• plaited-standards - Code conventions and standards
• plaited-ui-patterns - UI templates and styling
• plaited-web-patterns - Web component patterns

Source: plaited/plaited

License

ISC