Free Cursor rules, a lean CLAUDE.md template, Claude Code hooks, and a custom GPT, each shipped with notes on exactly how it was tested. Enter your email to download; you'll also get Agent Digest, the weekly AI digest for builders, every Thursday.
One email unlocks the download. Unsubscribe anytime.
5 .cursor/rules files that keep the agent from breaking working code. They enforce small diffs, tests, git checkpoints, and a pre-ship checklist.
A lean CLAUDE.md, a memory file and a handoff file, plus remember and /handoff skills and safe permission defaults.
Paste-ready GPT instructions, starters, and knowledge files. It turns a rough idea into a one-page spec plus copy-paste build prompts.
⬇ Download zipA PreToolUse hook that blocks commits and file writes containing API keys, private keys, or .env files. Comes with a 15-case test script.
Or grab all four in one zipall four at once after any signup.
Planner, builder, reviewer, and tester subagents run by a single /ship command, with install notes and test results.
Pro ($8/mo, coming soon) will include every template.
Cursor rules are .mdc files in a project's .cursor/rules folder that give Cursor's agent standing instructions. Vibe-to-Prod Rules is 5 files: core guardrails (always on), testing, git checkpoints, frontend safety (set to auto-attach to .tsx, .jsx, .vue, and .svelte files), and a manual pre-ship checklist. They push the agent toward small diffs, tests, and checkpoints so it stops breaking code that already works.
Yes. The Starter CLAUDE.md + Memory Kit is a 24-line CLAUDE.md that imports a memory file and a handoff file, plus remember and /handoff skills and a settings.json with allow and deny permission rules. It is deliberately short, because long CLAUDE.md files get ignored.
Every download includes a TESTING.md that lists what was checked, on what date, and what has not been tested yet. For example, the Secrets Guard Hook passes 15 of 15 cases in its included test script, and the Ship-It Subagent Team passes 57 automated format checks. Live-session tests in Cursor, Claude Code, and ChatGPT are listed separately as not yet done where that is the case.
It is a PreToolUse hook. Before Claude Code runs a git commit or writes a file, the hook scans for API keys (AWS, GitHub, OpenAI, Stripe), private keys, and .env files, and blocks the action with a reason if it finds one. A secrets-allow comment marks a known false positive.
Four Claude Code subagents (planner, builder, reviewer, tester) and a /ship command that takes a small feature from plan to reviewed, tested diff, looping back to the builder at most twice. Only the builder can edit code, and it never commits or pushes. You get an instant download after paying through Stripe, with install notes, an example run, and the test results.
Yes. Enter your email to download, which also subscribes you to Agent Digest, a free weekly AI digest for builders. You can unsubscribe anytime and keep the files.