A Markdown workflow kit that makes Cursor, Claude Code, Codex and Windsurf ask before they edit. Free, open source.
npx agentslice initInside AgentSlice
Keep the workflow plain, visible and close to the code. AgentSlice gives your AI coding tool a small path to follow before implementation starts — and the whole kit lives in your repo as Markdown you can read, fork and edit.
01
The kit lives in your repo as Markdown files. Prompts, rules, planning skills and workflow state are easy to read, change and commit — and easy to fork on GitHub.
02
Instead of wandering edits and context drift, the workflow asks for product intent, audience, constraints and what should not change.
03
AgentSlice narrows the next move into a small, reviewable slice and leaves a QA handoff before the agent continues.
Build only the empty state, add copy, then run QA before the next slice.
Open source
AgentSlice Core is MIT-licensed and lives on GitHub. Install with one command. Fork it. Edit it. Make it yours.
npx agentslice initFAQ
A public GitHub repo of Markdown files: one install prompt, workflow templates, planning skills and tool-specific rules. Run npx agentslice init in your project and the kit drops into your repo, ready to use.
I spent 3 months experimenting with building AI products and testing different workflows. This is the workflow that actually worked — intake questions, slice discipline, QA handoff, cross-tool consistency. The category norm in 2026 is open and forkable: awesome-cursorrules, BMAD, shadcn all proved it. So I'm making it available free for everyone.
Hooks break across tools and lock you into one editor. Plain Markdown travels with your repo, survives model and tool changes, and stays readable for the whole team. If you need hard enforcement, pair AgentSlice with Cursor's PLAN/ASK toggle or Claude Code's hooks — the workflow is portable on purpose.
Markdown is persuasion, not enforcement — and that's the honest tradeoff. What changes is the default: the workflow stays persistent across every turn, so your agent leans plan-first instead of edit-first. In practice that means fewer drifts, fewer surprises, fewer file edits you didn't ask for.
You can. AgentSlice ships more than rules: a workflow-state file the agent reads each turn, planning skills for slice-by-slice work, a QA-handoff template, and tool-specific rule files for Cursor, Claude Code, Codex and Windsurf. The workflow around the rules is what does the work.
Cursor, Claude Code, Codex, Windsurf and similar AI coding tools. The kit ships rules and skills tuned for each.
Yes — that's the point. MIT license, plain Markdown, open repo. Fork it, adapt prompts, rules and templates to your project, stack and team conventions. PRs welcome.