# INDEX.md

## Start Here
1. Read `AGENTS.md`.
2. Read `HARNESS.md`.
3. Read `00_開発共通/開発時に共通の前提情報.txt`.
4. Confirm task contract (Purpose/Target/Done).

## Priority Map
- Rules:
  - `AGENTS.md`
  - `HARNESS.md`
- Shared context:
  - `00_開発共通/開発時に共通の前提情報.txt`
- Docs:
  - `docs/` (specs, runbooks, decisions)
- Source:
  - project code directories
- State:
  - `state/tasks/` (task records)
  - `tests/golden/` (regression dataset)

## Work Sequence
1. Clarify scope from task contract.
2. Identify minimal files to read/edit.
3. Implement in small reversible steps.
4. Run verification (L1/L2/L3 as applicable).
5. Record results in `state/tasks/<task_id>.md`.
6. Update docs if behavior/spec changed.

## Input Template for Agents
Use this payload shape:
- Purpose: ...
- Target: ...
- Done: ...
- Constraints: ...
- Relevant files:
  - path/to/file1
  - path/to/file2
- Verification commands:
  - cmd1
  - cmd2

## Prompt Budget Rules
- Prefer path references over pasted code.
- Provide excerpts only when necessary.
- Keep assumptions explicit and short.
- If token pressure increases, summarize and continue.

## Verification Checklist
- L1 static checks executed.
- L2 focused tests executed.
- L3 scenario/regression checks executed or justified.
- Results captured with command + exit code + key logs.

## Definition of Done
- Requested behavior is implemented.
- Verification evidence is recorded.
- Docs/specs are updated if needed.
- Open risks and next actions are explicitly listed.

## Quick Commands (edit as needed)
```powershell
# Example placeholders
# npm run lint
# npm test
# npm run build
```