5.6 KiB
Phase 4: Harden + clean up (optional) - Discussion Log
Audit trail only. Do not use as input to planning, research, or execution agents. Decisions are captured in CONTEXT.md — this log preserves the alternatives considered.
Date: 2026-05-20 Phase: 04-harden-clean-up-optional Areas discussed: Phase 4 scope decision, Audit P1 polish selection, Visual polish selection, Alpha tester integration, ROADMAP backfill
Phase 4 scope decision
| Option | Description | Selected |
|---|---|---|
| Full Phase 4 | All 4 ROADMAP SC + meaningful subset of 12 deferred items. ~5-7 plans across 2-3 waves. Excludes rrweb v2 + programmatic SW-RAM (defer to v2). Est 4-8h. | ✓ |
| Focused Phase 4 (minimum acceptable) | 4 ROADMAP SC + 2-3 most-impactful deferred items. ~3-4 plans. Est 2-4h. | |
| Close milestone v1 at Phase 3 (skip Phase 4) | Re-distribute current build for testers; gather feedback over 1-2 weeks; schedule v1.1 as follow-up. Cleanest closure. | |
| Hybrid: dead-code cleanup + alpha rebuild | Quick 1-plan phase: SC #4 grep-clean + repackage. ~30-60 min. |
User's choice: Full Phase 4 (Recommended for v1 quality)
Notes: Captured in CONTEXT.md D-P4-01. Honors no-unilateral-scope-reduction memory + canonical ceremony preference. Phase 4 is the last hardening window before v1 release.
Audit P1 polish items — fix all three correctness items, or cherry-pick?
| Option | Description | Selected |
|---|---|---|
| All three (#11+#14+#15) | P1 #11 fetch URL extraction + #14 nav URL tracking + #15 rrweb timestamps semantics. Cohesive content-script polish. | ✓ |
| Just #11 fetch | Skip #14 + #15 (less critical). | |
| Cherry-pick by plan-time research | Defer decision to plan-time. |
User's choice: All three (#11+#14+#15) (Recommended)
Notes: Captured in CONTEXT.md D-P4-02. All three are correctness fixes in src/content/index.ts; small surgical scopes; high value-to-cost ratio.
Visual polish — include operator-perceptible UX items?
| Option | Description | Selected |
|---|---|---|
| Both: cursor visibility + dark-logo contrast | getDisplayMedia cursor:'always' (Plan 01-07 obs 2026-05-15; pointer-driven bug repro) + dark-logo (Plan 01-10 obs 2026-05-20). Small surgical fixes; operator-facing quality. | ✓ |
| Only cursor visibility | Cursor is functional; dark-logo is purely aesthetic. | |
| Defer both to v2 | Operator can still reproduce bugs without cursor; dark-logo is theme polish. |
User's choice: Both: cursor visibility + dark-logo contrast (Recommended)
Notes: Captured in CONTEXT.md D-P4-03. Both improve operator quality before alpha re-distribution.
Alpha tester feedback integration
| Option | Description | Selected |
|---|---|---|
| Proceed now; re-distribute after Phase 4 | Current alpha is pre-Phase-2/3; re-package fresh build after Phase 4 closes. Faster to v1. | (implied via user freeform) |
| Pause Phase 4 — wait for current alpha feedback first | Hold Phase 4 until testers report. Slower; risk of indefinite wait. | |
| Interleave: re-distribute mid-Phase-4 after Wave 1 | Re-package after Wave 1 (bug fixes + polish) lands; testers validate partial improvements. |
User's choice (freeform): "no no, if something i'll tell you"
Interpretation: Don't wait on alpha tester feedback; user handles operator-side signal out-of-band themselves. Phase 4 execution proceeds independently. After Phase 4 closes, fresh build re-packaged for testers as v1-final candidate.
Notes: Captured in CONTEXT.md D-P4-04. Faster to v1 close; honors user's operating mode.
ROADMAP backfill for Plans 01-08..01-13 — include in Phase 4 or defer?
| Option | Description | Selected |
|---|---|---|
| Include (docs hygiene) | Add missing per-plan rows to ROADMAP.md Phase 1 plan table. Small surgical docs polish; ~15 min. | ✓ |
| Defer to v2 (Recommended in option text — cosmetic) | ROADMAP rows are cosmetic; SUMMARYs + commits are authoritative. |
User's choice: Include (docs hygiene)
Notes: Captured in CONTEXT.md D-P4-05. User chose against the cosmetic-defer recommendation — keeps canonical reference document accurate.
Claude's Discretion
Captured in CONTEXT.md <decisions> "Claude's Discretion" subsection:
- Plan organization (6-8 plan split; suggested 8-plan layout with consolidation room for 6-7)
- Wave structure (3-4 waves; sequential where file-overlap)
- Harness assertion numbering (A33+ continuing Phase 3 sequence)
- Pre-checkpoint bundle gates (6/6 inventory standard)
- Tier-1 FORBIDDEN_HOOK_STRINGS (stay at 12 unless A33+ needs new gated symbols)
- CSP-safety mitigation for setimmediate polyfill (3 option paths; planner picks)
Deferred Ideas (out of Phase 4 scope)
All captured in CONTEXT.md <deferred> section:
- To v1.1 / v2 maintenance: rrweb v2 upgrade, programmatic SW-RAM, REQ-password-confidentiality v2, ROADMAP backfill for any other inline-tracked plans, alpha-tester findings integration (user-handled out-of-band)
- To v2 / SRV milestone (per SPEC §9): server upload, AI diagnostics, automatic ticketing, analytics dashboard, audio recording, DOM-event masking reactivation
Session efficiency notes
- User selected "Full Phase 4" — broadest scope per no-unilateral-scope-reduction memory
- 5 decision questions across 2 AskUserQuestion batches (1 initial scope + 1 follow-up 4-question batch)
- Total: 2 user-interaction turns
- Continues established efficiency pattern (Phase 2: 2-3 turns; Phase 3: 3 turns)
- Aligned with user's "minimum friction everywhere" + "canonical ceremony" preferences