klaussy

klaussy

Multi-agent repo boilerplate generator. One command to make any repo ready for Claude Code, Gemini CLI, Cursor, Codex, and GitHub Copilot — each gets the same conventions and the same workflow skills in its own native format.

Install

pip install klaussy-agents

Requires klaussy-repo-conventions (installed automatically).

Quick Start

cd your-repo
klaussy init

That's it. You'll be prompted for your base branch (auto-detects dev, main, etc.), then klaussy generates everything.

By default klaussy bootstraps all supported agents from the same conventions. To narrow to a subset, pass --agents:

klaussy init                                   # all agents (default)
klaussy init --agents claude                   # Claude Code only
klaussy init --agents claude,gemini,cursor     # a subset

See Multi-agent targets for what each agent gets.

What Gets Generated

klaussy discovers your repo's conventions once, then writes — for every selected agent (all five by default) — that agent's native conventions file, the workflow skills, stack-appropriate permissions, and hooks where the agent supports them. Narrow with --agents to emit only the agents you want.

# Per agent (each gets the workflow skills + a conventions file + permissions + hooks):

Claude Code   CLAUDE.md  .claude/rules/  .claude/skills/<repo>-<skill>/  .claude/settings.json  .claude/hooks/
Gemini CLI    GEMINI.md  .gemini/skills/<repo>-<skill>/  .gemini/settings.json  .gemini/hooks/  .geminiignore
Cursor        .cursor/rules/*.mdc  .cursor/skills/<repo>-<skill>/  .cursor/permissions.json  .cursor/hooks.json  .cursorignore
Codex         AGENTS.md  .agents/skills/<repo>-<skill>/  .codex/config.toml  .codex/hooks.json
Copilot       .github/copilot-instructions.md  .github/instructions/  .github/skills/<repo>-<skill>/  .github/hooks/

# Every skills/ directory holds the same namespaced set:
#   <repo>-{review, plan, debug, implement, refactor, test, fix, pr, commit, explain, new-worktree}

# Shared, once:
.github/PULL_REQUEST_TEMPLATE.md   # only if the repo doesn't already have one
.gitignore                         # appends klaussy output exclusions (pr-description.md, REVIEW_OUTPUT.md, plan.md)

See Multi-agent targets for the exact per-agent mapping (conventions, skill adaptation, secret exclusion, hook coverage), and the table below for what each skill does.

What each piece does

CLAUDE.md — Auto-detected conventions, architecture, commands, and pitfalls for your repo. As of 0.2.0 path-scoped rules are split out into individual files under .claude/rules/<glob-stem>.md (each with paths: frontmatter) so rules apply where they belong instead of as a flat list — CLAUDE.md itself holds the project-wide content. It's the conventions source each agent reads, emitted in that agent's native file (CLAUDE.md, GEMINI.md, AGENTS.md, .github/copilot-instructions.md, or .cursor/rules/).

settings.json — Auto-detects your stack (Python, Node, Go, Rust, Make) and sets tool permissions. Detects sensitive files (.env, *.pem, credentials*) and adds deny rules so the agent can't read them. Each agent gets this in its native form — see Permissions & secrets under Multi-agent targets.

Skills — Each repo gets a set of namespaced skills (<repo>-<skill>) so the agent auto-triggers them by description and they don't collide across repos. They're written into each selected agent's skills directory (.claude/skills/, .gemini/skills/, .cursor/skills/, .agents/skills/, .github/skills/). The bundled set is listed below; the canonical list lives in SKILL_NAMES in src/klaussy/skills.py. Prose-output skills (review, pr, commit, explain) share one humanization spec (HUMANIZE_BLOCK) so their output reads human — no em-dashes, filler openers, or chatbot scaffolding — across every agent. For a hard guarantee independent of how well the model complied, klaussy also ships a deterministic, code-preserving scrubber: the klaussy.humanize module and a klaussy humanize CLI (pipe a comment with printf '%s' "$c" | klaussy humanize, or scrub files with klaussy humanize FILE --write / --check). Both the prompt spec and the scrubber are faithful ports of klaussy-desktop's humanize-comment.js, so desktop and CI can pipe through this one canonical implementation instead of diverging.