Babel v0.2.0

What's Changed

• fix(chat): close chat-integration correctness findings (text tools, status routing, instructions) by @gthgomez in #376
• fix(deps): remove CLI install warnings for 0.1.4 by @gthgomez in #378
• fix(agent): report multi-file patch failures and rollback truthfully (M1) by @gthgomez in #381
• fix(desktop,remote): bind readiness and stop ownership to the started child (M2) by @gthgomez in #382
• fix(review): validate findings and close review audit gaps by @gthgomez in #379
• Fix TUI response integrity and simplify interactive presentation by @gthgomez in #380
• fix(chat): preserve intent, scoped recovery and terminal evidence by @gthgomez in #384
• fix: preserve executor JSON and recover Docker file inspections by @gthgomez in #383
• fix(chat): keep denied invocations out of verifier evidence by @gthgomez in #385
• fix(chat): preserve cancellation terminal journal parity by @gthgomez in #386
• fix(cli): bind frozen acceptance to trusted verifiers and per-criterion receipts by @gthgomez in #388
• fix(chat): preserve context recovery and terminal evidence by @gthgomez in #387
• fix(ci): batch TAP, scope probe, desktop, and unit shard prerequisites by @gthgomez in #389
• fix(windows): unit shard 2 cliEntryDispatch hooks and native temp paths by @gthgomez in #390
• Desktop product delivery + criterion binding (reconciled with main @ e470e9e) by @gthgomez in #377
• fix(chat): preserve task-owned edits across governed writes by @gthgomez in #392
• fix(babel-cli): recover compaction from complete request pressure by @gthgomez in #394
• feat(remote): Phase 1 pairing UI, admin boundary, and control leases by @gthgomez in #391
• fix(babel-cli): separate conversation recovery from repository capture by @gthgomez in #395
• feat(remote): shared Desktop controller ownership (stacked on Phase 1 pairing) by @gthgomez in #393
• fix(bridge): emergency approval.lookup test stage barrier (main regression) by @gthgomez in #396
• feat(bridge): Command Code own-CLI ACP harness adapter by @gthgomez in #397
• feat(bridge): recover OpenCode own-CLI ACP adapter (local review candidate) by @gthgomez in #400
• AGY offline static foundation (disabled) + first-publication prior-gate hardening by @gthgomez in #401
• fix(ci): exact-head PR resolve for workflow_run merge audit by @gthgomez in #403
• fix(babel-cli): enter the post-write repair budget only on repair evidence by @gthgomez in #398
• fix(babel-cli): close remaining coding-loop remediation findings (5-7 + guidance) by @gthgomez in #399
• fix(chat): retrieve owned output and harden recovery and cancellation by @gthgomez in #410
• fix(research-v2): post-review settlement, G-L1 readiness, and acquisition repairs by @gthgomez in #406
• fix(chat): align coding-loop discovery, output and verifier contracts by @gthgomez in #414
• fix(executor): handle Windows guard access and startup by @gthgomez in #415
• docs: make the first useful Babel task match the release and execution policy by @gthgomez in #416
• fix(desktop): match empty-state starters to the known project scope by @gthgomez in #417
• perf: reduce check latency and repeated agent work by @gthgomez in #418
• fix(bridge): Claude Code lifecycle, host registry, isolated policy proof by @gthgomez in #402
• test(trust-root): align product vs trusted readiness trigger expectations by @gthgomez in #421
• Performance remediation integration: intent, witness binding, protected inputs + retained repairs by @gthgomez in #419
• docs(agent): Slice A batch handoff and maintainer supplemental preflight diagnostic by @gthgomez in #422
• fix(tools): pre-publication AGENTS budget and surface checks by @gthgomez in #423
• fix(integration): qualify acceptance, answer retention and dispatch accounting by @gthgomez in #438
• Repair native Chat stopping and prepare frozen backend evaluation by @gthgomez in #425
• fix(P02): suppress repeat isolation shell denials at ChatEngine boundary by @gthgomez in #427
• Integrate P02 isolation + H4 artifact journey with P05 campaign budget wire (combined candidate) by @gthgomez in #428
• fix(agent): preserve UNVERIFIED substantive answer at turn cap (J08 P1) by @gthgomez in #434
• fix(agent): refresh preserved UNVERIFIED capture per mutation revision (J08 P1 successor) by @gthgomez in #435
• fix(runners): campaign pre-wire zero settlement and dispatch evidence by @gthgomez in #436
• Versioned task acceptance v2 provisional inferred admission (frozen v1 replay) by @gthgomez in #437
• feat(cli): opt-in ordinary_delivery completion profile by @gthgomez in #439
• J03: authoritative allowance stop, retry finalization, finite 429 accounting tests by @gthgomez in #440
• feat(cost): separate legacy campaign ceiling rates from usage estimates by @gthgomez in #441
• Task 2: narrow charge provenance and immutable priced settlement by @gthgomez in #442
• fix(chat): unlimited task dollar defaults — HOLD correction batch by @gthgomez in #443
• feat(authority): stage-1 trusted-host path-list declaration and pure verifier by @gthgomez in #430
• fix(ui): distinguish incomplete cost accounting in terminal titles by @gthgomez in #444
• feat(desktop): support explicit provider model routes by @gthgomez in #445
• fix: meter delegated inference against the parent campaign by @gthgomez in #446
• fix: retain unverified file edits when shell verification is unavailable by @gthgomez in #447
• fix(cli): preserve native Command Code inspection routing by @gthgomez in #448
• fix(authority): honor coordinated publication prohibitions by @gthgomez in #449
• fix(authority): distinguish explanatory publication quotes from action requests by @gthgomez in #452
• fix(chat): preserve blocking choice questions as user deferrals by @gthgomez in #451
• fix(cli): honor explicit artifact-saving intent and read-only mode by @gthgomez in #450
• fix(cli): treat literal rg search patterns as data by @gthgomez in #454
• fix(chat): preserve answer context and scope tool-repair guidance by @gthgomez in #453
• feat(runs): add immutable local batch review of run evidence by @gthgomez in #455
• fix(chat): leave whole-task wall deadlines unset by default by @gthgomez in #458
• fix(merge): qualify against authenticated live main by @gthgomez in #459
• fix(glob): preserve structured path counts in tool feedback by @gthgomez in #456
• fix(chat): retain honest unverified native edits when tests cannot launch by @gthgomez in #457
• feat(ci): add offline advisory evidence planner by @gthgomez in #460
• fix(agent): preserve verifier intent and file observation boundaries by @gthgomez in #461
• Repair worktree governance and native test-count claims by @gthgomez in #462
• fix(agent): guide test admission and credit executed verifiers by @gthgomez in #463
• fix: forward bounded merge-gate evidence wait by @gthgomez in #465
• fix: localize TypeScript Node test failures by @gthgomez in #464
• fix(agent): complete governed recovery and truthful verification delivery by @gthgomez in #468
• fix: clarify verifier feedback and recover empty answers by @gthgomez in #469
• Improve governed Python recovery and verifier evidence by @gthgomez in #470
• chore(release): prepare Babel 0.2.0 by @gthgomez in #471

Full Changelog: v0.1.3...v0.2.0