Built a Claude skill that generates .cursorrules as part of a full Spec Driven Development setup. Here’s what makes it different from a generic .cursorrules template:
It generates from your actual project context. The skill interviews you first (what the project does, your stack, your deployment target), then generates a .cursorrules file with the Universal Instruction Block filled in for your specific project. No {{PROJECT_NAME}} placeholders.
It generates matching configs for every other tool too. Same session, same interview answers → CLAUDE.md, .windsurfrules, .github/copilot-instructions.md, .aider.conf.yml. The Universal Instruction Block is byte-identical across all of them. If your team uses Cursor for some things and Copilot for others, they read the same mandate.
The block points every agent at your spec files:
MANDATORY BEFORE ANY ACTION:
0. If CONTEXT.md exists, read it first — it has session state
1. Read requirements.md in full
2. Read design.md in full
3. Read tasks.md — identify the next incomplete task
It also handles existing codebases. “I already have a codebase, no specs yet” triggers a retrofit workflow that reverse-engineers your spec from what you describe and generates the .cursorrules from that.
Looking for Cursor users specifically to test whether the generated .cursorrules files work without modification. The cross-AI consistency claim (byte-identical UIB across 4 config files) passed lab testing — I need it tested in the wild.
Repo: GitHub - FredAntB/Spec-Driven-Development · GitHub
Open an issue tagged [Beta] to volunteer.