AzuTear
e76d88e0c7
Import the existing Electron + React + TypeScript app as the version-control baseline before the scanner rework (C# input/capture sidecar, resolution-anchored layout profiles, OCR preprocessing, eval harness, rescan-merge, GOOD interop). Housekeeping in this commit: - Remove orphaned temp_inputhelper_block.ts (duplicate of the input-helper script). - Ignore .claude/scheduled_tasks.lock local session state. - Add .gitattributes to normalize line endings (LF in repo). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
64 lines
2.4 KiB
Markdown
64 lines
2.4 KiB
Markdown
# AGENTS.md
|
|
|
|
This file defines how AI coding agents should work in this repository.
|
|
|
|
## Core Principles
|
|
|
|
1. Correctness comes before speed.
|
|
2. Preserve the scanner-first product goal: reduce artifact decisions, guide hopping, and manual bookkeeping.
|
|
3. Prefer small, reviewable changes over broad rewrites.
|
|
4. Do not invent Genshin game data, API behavior, OCR confidence, or optimization rules without documenting the assumption.
|
|
5. Keep all game interaction read-only by default. Never add delete, feed, enhance, memory-read, hook, or game-modification behavior.
|
|
6. Keep user work and local app state intact.
|
|
7. Validate Electron main-process changes with a build and, when practical, a manual Smart Capture smoke test.
|
|
|
|
## Read Before Write
|
|
|
|
Before editing code, inspect the relevant source files and the project documents:
|
|
|
|
- `docs/PROJECT.md`
|
|
- `docs/ARCHITECTURE.md`
|
|
- `docs/CONVENTIONS.md`
|
|
- `docs/DECISIONS.md`
|
|
- `docs/CHECKLISTS.md`
|
|
|
|
Use `rg` or `rg --files` for search.
|
|
|
|
## Product Safety
|
|
|
|
The app may capture the Genshin window or screen and may later support optional input automation. Treat those areas as high risk.
|
|
|
|
- Capture must not require memory reads, hooks, injection, or game file modification.
|
|
- Input automation, if added, must be opt-in, reversible, whitelisted, and stoppable.
|
|
- The app must never delete, feed, enhance, or spend in-game resources.
|
|
- Uncertain OCR data must remain reviewable instead of being silently trusted.
|
|
|
|
## Implementation Expectations
|
|
|
|
- Keep Electron main-process code focused on OS integration, capture, and IPC.
|
|
- Keep React code focused on UI state and presentation.
|
|
- Keep parsing, scoring, and recommendation logic in `src/lib`.
|
|
- Keep domain types in `src/types`.
|
|
- Prefer deterministic parsers and data packages over hardcoded one-off exceptions.
|
|
- When OCR requires heuristics, record the limitation in docs or code comments.
|
|
|
|
## Validation
|
|
|
|
Run the strongest practical checks for the change:
|
|
|
|
- `npm run lint`
|
|
- `npm test`
|
|
- `npm run build`
|
|
|
|
For scanner changes, also manually verify Smart Capture against an open artifact detail view when possible.
|
|
|
|
## Definition Of Done
|
|
|
|
A task is done when:
|
|
|
|
- The requested behavior exists.
|
|
- The implementation follows the documented architecture and conventions.
|
|
- Relevant checks passed or skipped checks are explained.
|
|
- Scanner/OCR changes fail safely into review or uncertainty.
|
|
- The final summary names the changed areas and remaining risks.
|