Implement the Prometheus Chromium browser through Phase 22

.gitignore

@@ -0,0 +1,12 @@
+# Local Chromium/depot_tools workspaces and generated artifacts
+/.context/
+/.gclient
+/.gclient_entries
+/depot_tools/
+/out/
+/*.log
+/*.status
+/*.tmp
+/*.cache
+/chromium-*.log
+/chromium-*.status

.planning/MILESTONES.md

@@ -0,0 +1,52 @@
+# Milestones
+
+## v2.0 Chromium Engine Migration (Active: started 2026-06-17)
+
+**Planned phases:** 14 phases (Phases 11-24)
+**Scope:** Preserve v1 Falkon/Prometheus source and product decisions, retire Falkon as the active implementation, check out/build Chromium, and rebuild Prometheus on Chromium-native UI, WebUI, agent-control, MCP, runtime, supervision, extension, validation, and packaging primitives.
+
+**Key objectives:**
+
+- Preserve the Falkon-based v1 source reference as a recoverable bundle and patch series before deleting the active `falkon/` tree.
+- Carry forward the Prometheus/FSB product DNA: Prometheus name, Powered by FSB affiliation, DOM/native control thesis, MCP compatibility, supervision, secure vault/provider runtime, memory/site guides, compact chrome, warm themes, PM mark, FSB assets, Poppins/Space Mono, and Font Awesome Free.
+- Establish a clean upstream Chromium checkout/build baseline and record the exact revision.
+- Rebuild product identity, design system, chrome, WebUI surfaces, native agent service, page adapters, MCP bridge, FSB runtime, supervision, and extension policy on Chromium.
+- Complete parity, security, legal, visual, and release validation before calling the migration complete.
+- Keep Claude Code and other agent tools external through the stdio MCP bridge; do not bundle an in-browser terminal or third-party agent CLI runtime.
+- Consolidate remaining human visual UAT into a non-blocking Phase 24 release review while keeping automated acceptance gates mandatory.
+
+**Primary artifacts:**
+
+- `.planning/preservation/FALKON-TO-CHROMIUM-PRESERVATION.md`
+- `.planning/REQUIREMENTS.md`
+- `.planning/ROADMAP.md`
+- `.context/falkon-preservation/falkon-prometheus-v1.bundle`
+- `.context/falkon-preservation/patches/`
+
+**Status:** Phases 11-19 complete. External stdio MCP compatibility is verified
+without a bundled agent CLI, terminal, alternate engine, prompt provisioner, or
+third-party runtime. Phase 20 is current and ready for autonomous discussion
+and planning of the live FSB runtime, providers, vault, memory, logs,
+diagnostics, and control-panel data.
+
+---
+
+## v1.0 MVP (Shipped: 2026-06-17)
+
+**Phases completed:** 10 phases, 51 plans, 42 tasks
+**Timeline:** 2026-06-16 → 2026-06-17 (~2 days, 70 planning-repo commits)
+
+**Delivered:** An agent-native, FSB-powered desktop browser built on a rebranded Falkon/QtWebEngine fork — native agent control, MCP compatibility, FSB runtime parity, supervision, compact chrome, offline theming, and a start page — as a statically-verified local macOS validation build.
+
+**Key accomplishments:**
+
+- Built and legally inventoried a Falkon/QtWebEngine fork baseline, then rebranded all product-visible surfaces to Prometheus ("Powered by FSB") while preserving GPL/Qt/KDE legal notices (Phases 1-2).
+- Shipped a native agent control spine: command router, surface adapters, permissioned JavaScript execution, and DOM/page reads across web and browser-owned surfaces (Phase 3).
+- Restored FSB-compatible MCP control with browser-owned agent identity, tab ownership, concurrency caps, typed recovery errors, and implicit visual-session fields (Phase 4).
+- Rebuilt FSB runtime surfaces natively — agent side panel, multi-provider configuration, autopilot, logs/diagnostics, memory/site guides, and a secure vault — plus PhantomStream-style supervision, safe pairing, and a validated local macOS release package (Phases 5-6).
+- Added compact Safari-style chrome with advanced tab management (groups, overview/search, quick switch, unload/restore) and a native FSB-Plus control center backed by a release-gated feature parity matrix (Phases 7-8).
+- Delivered an offline Font Awesome Free icon system with Poppins/Space Mono brand fonts, warm dark/light themes, a real PM brand mark, and a Prometheus start page with omnibox/Ask FSB handoff, favorites, and suggested automations (Phases 9-10).
+
+**Known deferred items at close:** 3 verification files (26 human-visual UAT items across phases 08/09/10) carried forward to the next milestone — see STATE.md Deferred Items. The internal Falkon → Prometheus symbol rename and a notarized public macOS release also remain deferred.
+
+---

.planning/RETROSPECTIVE.md

@@ -0,0 +1,61 @@
+# Project Retrospective
+
+*A living document updated after each milestone. Lessons feed forward into future planning.*
+
+## Milestone: v1.0 — MVP
+
+**Shipped:** 2026-06-17
+**Phases:** 10 | **Plans:** 51 | **Sessions:** not tracked precisely (~2 calendar days, 2026-06-16 → 2026-06-17)
+
+### What Was Built
+- A rebranded, legally-inventoried Falkon/QtWebEngine fork ("Prometheus", Powered by FSB) with preserved GPL/Qt/KDE notices.
+- A native agent control spine (command router, surface adapters, permissioned JS execution, DOM/page reads) and an FSB-compatible MCP bridge with multi-agent ownership, typed errors, and visual-session fields.
+- Native FSB runtime surfaces (side panel, providers, autopilot, logs, memory/site guides, secure vault), PhantomStream-style supervision, safe pairing, and a local macOS validation package.
+- Compact Safari-style chrome with advanced tab management, a native FSB-Plus control center with a release-gated parity matrix, an offline icon/font/theme system, and a Prometheus start page.
+
+### What Worked
+- The GSD phase pipeline (discuss → plan → execute → verify → code review → UI audit) drove 51 plans across 10 phases with consistent atomic commits.
+- Tracking the heavy upstream browser code as a separate `falkon/` checkout kept the GPL codebase out of the planning repo while GSD tracked plans and summaries cleanly.
+- Static verification gates (177/177 build-link check, parity gate, code review) caught most issues without needing a live build.
+
+### What Was Inefficient
+- Requirements traceability drifted: phases 8-10 were inserted mid-milestone, but their requirement statuses stayed "Planned" and checkboxes were left inconsistent — all 20 had to be reconciled at milestone close.
+- 26 verification items could only be confirmed by a human running the binary; they accumulated unverified across phases 08/09/10 and shipped as deferred items rather than being cleared, so v1.0 closed without an end-to-end human run.
+- Milestone-close auto-extraction pulled garbage "accomplishments" from SUMMARY.md files ("Status:", code-review fragments) and had to be rewritten by hand.
+- Scope grew: phases 7-10 were added after the original six-phase release-hardening milestone, expanding v1.0 well beyond its initial definition.
+
+### Patterns Established
+- Maintain a release-gated FSB feature parity matrix as the definition of "done" for parity categories.
+- Offline-bundle all UI assets (icons, fonts) in Qt resources with attribution, and gate releases on their presence.
+- Mark verification items that genuinely require a human/binary as `human_needed` so they don't block code-complete but stay tracked as deferred.
+
+### Key Lessons
+1. Update requirement traceability statuses when a phase completes, not at milestone close — mid-milestone phase insertions otherwise leave the table inconsistent.
+2. For UI-heavy work that can't be grep-verified, schedule a dedicated human-visual verification pass before declaring the milestone shippable, or explicitly accept and defer the gap (as done here).
+3. Auto-extracted milestone accomplishments are unreliable across heterogeneous SUMMARY.md formats — always curate them.
+
+### Cost Observations
+- Model profile: `balanced` (per `.planning/config.json`); exact model mix not tracked.
+- Velocity (from STATE.md metrics): ~7 min average per plan; heavy parallelization across waves.
+- Notable: code-complete reached fast, but human-visual verification debt was traded off to get there.
+
+---
+
+## Cross-Milestone Trends
+
+### Process Evolution
+
+| Milestone | Sessions | Phases | Key Change |
+|-----------|----------|--------|------------|
+| v1.0 | n/a | 10 | Established the GSD pipeline; scope grew from 6 to 10 phases mid-milestone |
+
+### Cumulative Quality
+
+| Milestone | Tests | Coverage | Zero-Dep Additions |
+|-----------|-------|----------|-------------------|
+| v1.0 | static + Qt focused tests | n/a | offline icon/font assets bundled (no new network deps) |
+
+### Top Lessons (Verified Across Milestones)
+
+1. Keep requirement traceability current per-phase; reconciling at close is error-prone.
+2. UI verification that needs a human must be scheduled deliberately, not deferred indefinitely.