Keep it simple, stupid.

A rulebook for focused changes, safer decisions, and stopping a two-file change from becoming a six-layer architecture.

See how to use it

Practical rules for AI coding agents

No framework required. No custom index. No “enterprise-ready” abstraction for a button.

Less nonsense. More working software.

Less unnecessary code.

Reuse what fits. Add only what the task needs. Preserve existing behavior and user work; ask before destructive changes.

Less guessing.

Start with the context you provided. Follow relevant files and dependencies, then expand the search only when the task needs it.

Less AI slop.

Follow the product, brand, and existing design system. Use Impeccable for UI/UX work, then critique the result instead of repeating a template.

More proof.

Require local checks for new or changed API behavior. Report what actually passed, what could not be tested, and what remains uncertain.

Clear rules. Focused guides.

Global rules, eight task-specific guides, and the bundled Impeccable UI/UX skill. Read what the task needs; keep the global rules in scope.

Using Claude Code, Gemini CLI, or another coding agent?
You may need to rename AGENTS.md to the instruction filename expected by your tool (for example, CLAUDE.md or GEMINI.md).

Inside the ZIP

as-kiss.zip
AGENTS.md
README.md
agent-guides/
├── engineering.md
├── backend.md
├── frontend.md
├── uiux-ds.md
├── testing.md
├── code-style.md
├── payments.md
└── access.md
.agents/skills/impeccable/
├── SKILL.md
├── agents/
├── reference/
└── scripts/

Small scope. Clear boundaries.

Inspect the supplied context first, preserve user work, and verify the changed behavior. Global rules stay in scope even when only a few guide sections are needed.

Read the bundle overview · Read the global rules

Review access before you copy.

This bundle has env_file_read: on. It permits task-relevant reading of project-local environment files. It does not allow secret exposure or environment-file edits. Review access.md.

  • Engineering, backend, frontend, and scoped code formatting.
  • UI/UX routing through the bundled Impeccable skill.
  • Local API verification and protection of existing user work.
  • Payment safety guidance for new gateways and new payment flows.
  • Project-local access settings for environment-file reading.
  • Explicit approval for destructive operations and real external actions. No access to remotely deployed project environments.

Download. Review. Add to your project.

Keep the folder structure intact and review the access settings. Merge with existing project instructions before asking your agent to use them.

  1. Download the ZIP.

    Download as-kiss.zip, containing the files listed above.

  2. Extract it.

    You should see AGENTS.md, README.md, agent-guides/, and the hidden .agents/ folder.

  3. Review and merge into your project root.

    Place the rules where they should apply: at the shared root for a monorepo, or the app root for a single app. Keep agent-guides/ and .agents/ alongside the instruction file. Merge existing instructions and skills rather than overwriting them; use the bundled README as reference.

    Example layouts below. Highlighted entries are the files and folders to add or merge.

    Single app — alongside your app files
    my-app/
    ├── AGENTS.md
    ├── agent-guides/
    ├── .agents/
    │   └── skills/
    │       └── impeccable/
    ├── src/
    ├── package.json
    └── README.md
    Monorepo — above the apps that share the rules
    my-workspace/
    ├── AGENTS.md
    ├── agent-guides/
    ├── .agents/
    │   └── skills/
    │       └── impeccable/
    ├── apps/
    │   ├── web/
    │   └── api/
    └── README.md

    Only one app needs the rules? Put the same three highlighted entries inside that app’s root instead. Keep your existing project README; the bundle README is installation reference.

  4. Rename only if your tool requires it.

    Some agents expect names like CLAUDE.md or GEMINI.md. Keep agent-guides/ and .agents/ beside it.

  5. Let routing guide the task.

    Have your agent read the global rules first, then the relevant guide sections and required references. UI/UX work uses the bundled Impeccable skill; its launcher may download its runtime on first use.

Rules guide the work. Evidence earns trust.

  • Review the agent’s changes and verification results.
  • Keep reading and implementation scoped to the actual task.
  • Use disposable local data or provider sandboxes for testing.
  • Keep secrets, user work, and permission boundaries protected.
  • Treat these as instructions, not automatic enforcement or a guarantee of correct output.

Your agent already knows how to code.

Give it rules for what to change, when to ask, and how to show the work was verified.

See how to use it