prompticorn

A unified, tool-agnostic prompt architecture for managing AI coding-assistant configurations across 17 assistants.

Version: the published version is dynamic. pyproject.toml declares dynamic = ["version"] sourced from prompticorn/__about__.py, and CI/CD injects the real MAJOR.MINOR.PATCH at build time. Local and editable installs report 0.0.0.dev0. Check your installed version with pip show prompticorn.

Define your project's agents, conventions, and personas once, then generate the right config for whichever assistant your team uses:

• Kilo Code — IDE (.kilo/agents/) and CLI (.opencode/agents/)
• Cline — .clinerules
• Claude — .claude/ directory plus CLAUDE.md
• Cursor — .cursor/rules/ plus .cursorrules
• GitHub Copilot — .github/copilot-instructions.md
• GitHub Copilot Chat (VS Code) — .github/agents/ custom agents, .github/prompts/ slash commands, and applyTo .github/instructions/
• Roo Code — .roomodes custom modes plus .roo/ rules, skills, and commands
• JetBrains Junie (CLI) — .junie/ agents, skills, and commands plus AGENTS.md
• Zed — .agents/skills/ (agents-as-skills) plus AGENTS.md instructions
• Gemini CLI — .gemini/ agents, skills, and TOML commands plus AGENTS.md
• Amazon Q (CLI) — .amazonq/ JSON agents, rules/ conventions, and prompts
• Windsurf / Cascade — .windsurf/ skills, glob-scoped rules/, and workflows
• Continue.dev — .continue/ glob-scoped rules/ and invokable prompts/
• Aider — CONVENTIONS.md plus .aider.conf.yml (conventions only)
• OpenAI Codex (CLI) — AGENTS.md plus .agents/skills/ and .codex/config.toml

What's in the library

• 25 primary agents (architect, backend, frontend, code, test, debug, security, devops, and more)
• ~100 workflows in minimal and verbose variants
• ~95 specialized skills
• 29 languages with first-class conventions (prompticorn/agents/core/conventions-*.md)

Install

pip install prompticorn
# or
uv add prompticorn

This installs the prompticorn CLI command.

Quick Start

cd your-project
prompticorn init

init is interactive: it asks which assistant to configure, your repository type, prompt variant, personas, language-specific settings, and a set of project questions (database, ORM, error-handling pattern, commit style, PR-size limit, deploy target, and source-tree layout). It then writes .prompticorn/.prompticorn.yaml and generates the assistant's config files.

See docs/QUICKSTART.md for the full walkthrough.

Key Features

• Unified IR system — define agents once, generate for every supported tool.
• 15 production builders — Kilo, Cline, Claude, Cursor, Copilot, Copilot Chat, Roo Code, Junie, Zed, Gemini CLI, Amazon Q, Windsurf, Continue.dev, Aider, Codex.
• Minimal / verbose variants — trade tokens for detail at build time.
• Persona-based filtering — pick your team's roles and only relevant agents are generated.
• Spec-driven conventions — your language, runtime, package manager, test framework, linter, formatter, coverage targets, and project settings are baked into the generated conventions.
• Per-language source layouts — the core convention renders each language's standard source tree; flat is the default and src is selectable.
• Auto-discovery registry — agents are discovered from the bundled agents/ tree; no manual registration.

Commands

Documentation

• docs/QUICKSTART.md — 5-minute getting-started guide
• docs/INDEX.md — documentation navigation hub
• docs/PERSONAS.md — persona-based filtering

Development

git clone https://github.com/snoodleboot-io/prompticorn.git
cd prompticorn

# Install in editable mode (reports version 0.0.0.dev0)
uv pip install -e .

# Run tests with coverage (target ~85%)
uv run pytest

# Mutation testing
uv run mutmut run

# Lint, format, and type-check
uv run ruff check .
uv run ruff format .
uv run pyright