Skip to content

AGENTS.md

Latest commit

 

History

History
97 lines (70 loc) · 3.73 KB

AGENTS.md

File metadata and controls

97 lines (70 loc) · 3.73 KB

AGENTS.md

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

Core Principles

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.

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.
  • No any. No else (use early returns). No _ (wrap/handle all errors).
  • Descriptive names. Comments explain "why" (intent), not "what" (syntax).
  • 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. 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.

Repo Workflow

1. Repo Specifics

Vertical slices. Small writes. One artifact per PR.

  • Work in vertical slices (e.g., one API route + Service + Repo + Svelte UI component). Never build entire layers at once.
  • When using the write/edit tool, write small blocks — there is a bug that causes timeouts on large writes.
  • Every PR/session produces exactly one primary deliverable (one doc or one vertical code slice).
  • Every PR, commit, and terminal command must be tagged: [unit: unit-name].

2. Communication & State

Terse. No fluff. Git is the source of truth.

  • Use constraint-based language. No "I can help with that" or "Sure thing."
  • Read git status, git log, and PR comments to determine current state.
  • Every response must end with a "Files Affected" list using absolute repository paths.

3. QA & Execution Loop

Tests pass or you're not done. Auto-fix failures.

  • Task = "Done" only if make test passes.
  • If make test fails, loop-fix immediately. Do not request permission.

4. Project Directory Structure

Know where things live. Don't guess.

  • /backend: Go source code.
  • /frontend: SvelteKit source code.
  • /design/README.md: The global repo context.
  • /design/units/{unit-name}: All planning and architecture docs.
  • /design/units/README.md: The global Unit Index (Status: Discovery, Research, Design, Implementation, QA).

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.