Skip to content

AGENTS.md

Latest commit

 

History

History
83 lines (57 loc) · 3.98 KB

AGENTS.md

File metadata and controls

83 lines (57 loc) · 3.98 KB
Raw

AGENTS.md

Behavioral guidelines to reduce common LLM coding mistakes. Merge with project-specific instructions as needed.

Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

Before implementing:

  • State your assumptions explicitly. If uncertain, ask.
  • If multiple interpretations exist, present them - don't pick silently.
  • If a simpler approach exists, say so. Push back when warranted.
  • If something is unclear, stop. Name what's confusing. Ask.
  • Design System: Always check Apple's Human Interface Guidelines for design patterns or best practices (https://developer.apple.com/design/human-interface-guidelines/)
  • Subagents: Use subagents whenever useful and especially when using a top-tier model (e.g. Sol or Opus), so the subagents can run on mid-tier models (e.g. Terra or Sonnet) to save tokens/usage and work simultaneously. If something is very easy, even take low-tier models for subagents (e.g. Luna or Haiku).

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

  • No features beyond what was asked.
  • No abstractions for single-use code.
  • No "flexibility" or "configurability" that wasn't requested.
  • No error handling for impossible scenarios.
  • If you write 200 lines and it could be 50, rewrite it.

Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

When editing existing code:

  • Don't "improve" adjacent code, comments, or formatting.
  • Don't refactor things that aren't broken.
  • Match existing style, even if you'd do it differently.
  • If you notice unrelated dead code, mention it - don't delete it.

When your changes create orphans:

  • Remove imports/variables/functions that YOUR changes made unused.
  • Don't remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user's request.

4. Documentation Is Part of the Change

Read the docs before you start. Update them before you commit.

Read first — before any work:

  • Read README.md and STRUCTURE.md at the start of every task, before exploring the codebase and before editing anything. STRUCTURE.md is the project map; README.md is the user-facing contract. Don't skip either because the task looks small.
  • Treat both as possibly stale. If they contradict the code, the code wins — and fixing the doc becomes part of the task.

Update before any commit or push — required, not optional cleanup:

  • Re-check both files against your changes and update whatever is now wrong.
  • Update README.md whenever a change affects setup, usage, behavior, requirements, permissions, versioning, or anything a user or maintainer would reasonably expect to find there.
  • Update STRUCTURE.md whenever project structure, architecture, build configuration, dependencies, or version changes, so future chats have a current repository reference.
  • If neither file needs a change, say so explicitly instead of silently skipping the check.

AGENTS.md and CLAUDE.md are kept byte-identical. Any edit to one must be mirrored to the other in the same commit.

5. Goal-Driven Execution

Define success criteria. Loop until verified.

Transform tasks into verifiable goals:

  • "Add validation" → "Write tests for invalid inputs, then make them pass"
  • "Fix the bug" → "Write a test that reproduces it, then make it pass"
  • "Refactor X" → "Ensure tests pass before and after"

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.

These guidelines are working if: fewer unnecessary changes in diffs, fewer rewrites due to overcomplication, and clarifying questions come before implementation rather than after mistakes.