DonnyClaude adds a durable planning layer and deterministic verification gates to Claude Code so the model can't mark work done until tests and checks pass. Install with npx donnyclaude.
Key Takeaways
- DonnyClaude adds a durable planning layer and deterministic verification gates to Claude Code so the model can't mark work done until tests and checks pass.
- Install with npx donnyclaude.
What Changed — The specific update, with version/date if available
DonnyClaude (v0.1) is a verified workflow engine for Claude Code, released as open source on GitHub. It addresses a fundamental weakness in AI coding tools: the model can call a task "done" before tests, requirements, or security checks actually pass.
While Claude Code is excellent at generating code, it struggles with long projects. Context gets messy after /clear, decisions get lost, and one overloaded prompt tries to plan, code, review, and verify itself. DonnyClaude wraps Claude Code in a durable planning layer and a deterministic verification loop.
What It Means For You — Concrete impact on daily Claude Code usage
DonnyClaude changes how you use Claude Code for multi-step projects. Instead of trusting the model to declare success, it enforces a workflow:
-
Writes project state to
.planning/— Work survives/clearand new sessions. Decisions are durable, not lost in chat history. - Breaks large work into explicit phases — With requirement and dependency gates between each phase.
- Routes work to scoped subagents — Planners, implementers, reviewers, and verifiers each handle their role.
- Refuses to mark work complete — Until engine-backed checks pass. No more "looks good" replacing actual tests.
Most Claude Code setups are prompt packs: they change what the model is told, then still trust it to declare success. DonnyClaude is built around one idea: the model generates the work, but completion is decided by durable state, skeptical review, and deterministic gates.
Try It Now — Commands, config, or prompts to take advantage of this
Install DonnyClaude:
npx donnyclaude
This installs the Donny engine, skills, agents, coding rules for 13 languages, project hooks, and a curated MCP setup (Context7 for live docs, Playwright for browser). Then Claude itself walks you through your first project.
Key commands:
-
npx donnyclaude --dry-run— Preview what install would change (no writes) -
npx donnyclaude doctor— Check installation health -
npx donnyclaude diff— Drift between installed files and this version -
npx donnyclaude uninstall— Remove every DonnyClaude-owned file; yours stay
Inside Claude Code:
/donny-init
This triggers research → requirements generation.
What gets installed:
-
~/.claude/skills/,agents/,commands/,hooks/,rules/— The toolkit -
~/.claude/donny/— The workflow engine -
~/.claude/bin/,cco-memory/,scrapers/,docs/,statusline.py— Support tools -
~/.claude/CLAUDE.md— Full operating guide (or idempotent standards block appended to existing file) -
~/.claude/settings.json— Merged with your original backed up - MCP registrations for Context7 and Playwright
Existing settings are preserved, not clobbered. Uninstall removes only files this package ships, strips just the managed block from CLAUDE.md, leaves settings.json in place, and never touches anything you created yourself.
When To Use It
DonnyClaude shines for any project where:
- You need to survive
/clearwithout losing decisions - You want to enforce verification gates (tests, security checks, requirement coverage)
- You're working on multi-phase projects that need planning, implementation, review, and verification
- You want a deterministic "done" check instead of trusting the model to grade its own output
For quick one-off scripts or single-file edits, DonnyClaude is overkill. But for any project where "looks good" isn't good enough, it's the missing layer.
Source: github.com
Originally published on gentic.news
Top comments (0)