Build a high-confidence map of the TaxHacker product and codebase (architecture, runtime flows, data model, feature surface, deployment, and current work-in-progress) so we can continue implementation work with minimal ramp-up.
Phase 2
- Understand user intent
- Identify constraints and requirements
- Document findings in findings.md
- Status: complete
- Map folder and module structure
- Identify key runtime entry points and app flows
- Capture dependencies, scripts, and environment expectations
- Status: in_progress
- Map core user journeys and feature modules
- Map data model (Prisma + storage)
- Identify integration points (AI providers, auth, OCR, currency)
- Status: pending
- Capture git status and active change areas
- Identify likely unstable areas and technical debt signals
- Capture missing docs and unanswered questions
- Status: pending
- Produce concise architecture summary
- Produce prioritized next actions to continue work
- Confirm context files are complete and current
- Status: pending
- Which modules are critical for day-to-day feature work and bug fixing?
- What is already changed locally versus baseline upstream behavior?
- Which assumptions (env vars, external services, binaries) are required to run and test confidently?
| Decision | Rationale |
|---|---|
| Use planning files in project root before deep exploration | Required by planning skill; keeps context persistent across long sessions |
| Preserve existing dirty worktree and avoid reverting user changes | Prevents accidental loss of active in-progress work |
| Error | Attempt | Resolution |
|---|---|---|
| session-catchup.py produced no output | 1 | Proceeded with manual context initialization and git diff baseline |
- Update phase status as discovery progresses.
- Log any failed command or dead-end search before trying a new approach.