Plan & PRD Review — Issue Tracker¶
Last updated: 2026-06-15
PRD version: v0.4
Related: PRD · Competitive landscape
Summary¶
| Status | Count |
|---|---|
| ✅ Resolved | 18 |
| ⏳ Open | 0 |
All plan review issues addressed in docs. Ready for PRD v0.4 approval gate.
Critical issues¶
1. Positioning says "proxy" — understates product¶
Status: ✅ Resolved (PRD v0.3+)
Product principles section; "local execution router" / "local cost optimizer" language throughout.
2. MVP too large for solo builder¶
Status: ✅ Resolved (PRD v0.3+)
Tracer bullet Phase A defined. Explicit defer list. Phase A.1 for setup + L6.
3. Router quality hand-waved¶
Status: ✅ Resolved
→ routing-spec.md: classifier inputs, heuristics, confidence scoring, escalation, 20 golden prompts.
4. Lt tier hardest to build¶
Status: ✅ Resolved
- Phase B.0: git, formatter, linter only
- Phase B.1: IntelliJ + destructive confirmation
- PRD user story #14
- routing-spec Lt patterns with
X-Daari-Confirm-Tool
5. Agent tool-call compatibility minefield¶
Status: ✅ Resolved
→ ADR-0004
6. Claude Code may not use OpenAI-compat¶
Status: ✅ Resolved
Client support matrix in PRD; Claude Code Phase B; Anthropic gateway Phase C2; setup-spec documents limitation.
7. Success metrics lack baseline¶
Status: ✅ Resolved (PRD v0.3+)
$0 tier rate, local AI rate, frontier rate, cost saved vs all-L6 baseline.
8. Open source strategy underspecified¶
Status: ✅ Resolved (PRD v0.3+)
OSS & privacy commitments section + ADR-0006 telemetry off by default.
Medium issues¶
| # | Issue | Status | Resolution |
|---|---|---|---|
| 9 | Language undecided | ✅ | ADR-0005 — Python 3.12 |
| 10 | PRD-PLAN status stale | ✅ | Updated to Phase 5 / v0.4 |
| 11 | No glossary | ✅ | glossary.md |
| 12 | Phase C too bundled | ✅ | Split C1 (MCP, profiles) / C2 (Anthropic, IDE) |
| 13 | daari.dev/install fake | ✅ | setup-spec.md — ./install.sh only for MVP |
| 14 | Eval deferred to v1 | ✅ | 20 golden prompts in routing-spec; Phase A eval |
| 15 | Setup reversibility | ✅ | setup-spec.md — backup + --undo |
Low issues¶
| # | Issue | Status | Resolution |
|---|---|---|---|
| 16 | Approval checklist incomplete | ✅ | All ADRs 0001–0006 + specs in PRD |
| 17 | Discovery frontier TBD | ✅ | Fixed in 02-discovery.md |
| 18 | No threat model | ✅ | ADR-0006 |
Next steps (post-review)¶
- You approve PRD v0.4 — check approval section in PRD
- Implementation plan — Phase A tracer bullet tasks
- Optional: challenge Python choice (OD-2) before coding starts
What's still hard (known, not blockers)¶
These are accepted risks, not open doc issues:
| Risk | Mitigation in docs |
|---|---|
| Local models weak on agent tool_calls | ADR-0004 → L6 escalation |
| Lt intent matching imperfect | Explicit regex patterns only; confirmation for destructive |
| Solo builder velocity | Tracer bullet MVP scope |