# AGENTS.md

## Project Context

**Goal**: Automate the e-filing process for legal documents.

**Approach**: Research-driven development. We first study how filing works today — both the manual process and competitor solutions — before building an automated solution.

## Directory Structure

| Path | Purpose |
|------|---------|
| `videos/` | Reference videos: manual filing walkthroughs + competitor demos |
| `docs/research/` | Research notes, video transcripts, skill-based analyses |
| `app/` | Production codebase (Python, managed with uv) |

## Key Guidelines for AI Agents

### Phase 1 — Research
- Analyze videos in `videos/` to understand the manual e-filing flow
- Compare with competitor automated solutions
- Document findings as structured notes in `docs/research/`
- Skills may be used to extract insights, frame captures, or transcripts from the video content
- Each research artifact should be self-contained and reference its source video

### Phase 2 — Build
- All implementation lives in `app/`
- Use **Python** as the language
- Use **uv** for dependency management (`pyproject.toml`, not `requirements.txt`)
- Virtual environments are managed by uv (no manual `venv`)
- Keep research docs separate from code — `docs/research/` is for knowledge, `app/` is for implementation

### Coding Standards
- Prefer clarity and readability over cleverness
- Reference the research when designing features
- Document assumptions and decisions in the code or as docstrings
- Use structured output formats (JSON, markdown) where applicable