@pellux/goodvibes-agent 1.1.6 → 1.2.0

package/CHANGELOG.md

@@ -2,6 +2,182 @@
 
 Product-facing release notes for GoodVibes Agent.
 
+## 1.2.0 - 2026-06-07
+
+- Promoted GoodVibes Agent to the stable 1.0.x operator product surface: the fullscreen Agent workspace is the primary TUI, with setup, provider/model routing, status, compatibility, and doctor flows shaped around personal operator use instead of copied host lifecycle controls.
+- Completed the Agent-local behavior system for day-one operation: VIBE.md personality, project context files, local memory posture, prompt-active recall, vector/embedding health, notes, personas, skills, skill bundles, routines, starter profiles, discovery/import flows, review/stale/delete controls, and secret-looking content rejection all stay under Agent ownership.
+- Completed isolated Agent Knowledge coverage across CLI, slash commands, workspace panels, connector/source/node/issue views, URL/file/browser/connector ingest, semantic ask/search, issue review, packet/explain previews, consolidation, reindex, and connected-host `/api/goodvibes-agent/knowledge/*` routes without fallback to default or non-Agent knowledge surfaces.
+- Completed connected-host operator integration without taking host lifecycle ownership: compatibility/status checks, authenticated health and model routes, channel readiness, provider-account posture, approvals, automation snapshots, schedules, work plans, media/voice readiness, pairing, and explicit public-route diagnostics are all visible from Agent.
+- Completed explicit side-effect boundaries for personal operation: channel sends, notifications, autonomous schedule creation, schedule enable/disable/delete, routine schedule promotion, reminders, subscription/auth actions, memory bundle imports/exports, support bundles, MCP configuration, profile changes, and build delegation require explicit confirmation where they mutate state or call external routes.
+- Added first-class user-task route planning: `route action:"plan|status"` and lower-level `agent_harness mode:"route_decision"` turn a plain request into the preferred visible Agent route, alternatives, missing fields, confirmation boundary, workspace matches, and harness mode matches without running tools or creating hidden work, including screenshot/browser-control requests that now route to `computer action:"plan"`.
+- Host/daemon health, doctor, readiness, service, and compatibility requests now route to `host action:"status"` first before setup repair or service lifecycle effects.
+- Normal settings/configuration requests now route to `settings action:"list"` first so setting changes start with Agent-owned discovery and explicit confirmation boundaries.
+- Direct reminders, schedules, cron, and schedule lifecycle requests now route to `schedule action:"list"` first, while broader ongoing work remains on autonomy intake.
+- Plain command-shaped background work now routes to `execution action:"processes"` and the first-class `terminal`/`process` UX, while scheduled or watcher-like background work stays on autonomy intake.
+- Interactive terminal, PTY, stdin, and sudo requests now route to `execution action:"process_capabilities"` first so users see current support, setup posture, and confirmation boundaries before any hidden process start or credential effect.
+- External memory-provider, backend, cross-session sync, import/export, and named-provider requests now route to `memory action:"provider"` or the external provider checklist before Agent promises provider writes, sync, credentials, or import/export effects; provider detail also exposes next-route packets, missing setup/status/read/write/sync checks, and required receipt fields for SDK/daemon records Agent can consume.
+- Model provider, local-cookbook, local server smoke, and route-fit requests now route to `models action:"provider|local|smoke|route"` before Agent attempts credential, smoke, benchmark, or route-change effects.
+- Browser-backed research runner requests now route to `research action:"runner"`, and visual research report rendering requests route to `research action:"plan"` plus report artifacts before Agent claims browser/PWA rendering readiness.
+- Voice workflow, TTS-provider, browser cockpit, and PWA requests now route to `device action:"voice|provider"` or `computer action:"browser"` before Agent attempts capture, playback, picker, or browser-open effects.
+- Personal Ops briefing, saved queue, fresh inbox/calendar read, and connector setup requests now route to `personal_ops action:"briefing|queue|intake|lane"` before Agent attempts live provider reads or effects.
+- Channel setup, triage, delivery receipt, and send requests now route to `channels action:"setup|triage|deliveries|channel"` before Agent attempts confirmed external delivery.
+- Plain file undo/redo/recovery requests now route to `execution action:"recovery"` so users inspect available snapshots before confirming a local file mutation.
+- Media generation requests now route to provider readiness and confirmed `agent_media_generate` saved-artifact output instead of inline bytes or silent Knowledge promotion.
+- Permission posture, security finding, and blocked-action questions now route to `security action:"status|finding|explain"` so users get active permission state, exact redacted findings, or read-only policy preflight without knowing security harness mode names.
+- Support-bundle, saved-session/bookmark, and release/audit evidence requests now route to first-class read-only `support action:"status|bundle"`, `sessions action:"list|get"`, and `audit action:"readiness|evidence|item|artifact"` before bundle export/import/share, session lifecycle, or audit drill-in effects.
+- Added first-class setup repair decisions: `setup action:"repair"` and lower-level `agent_harness mode:"setup_repair"` choose the next safe token repair, connected-host status, services.status receipt, user-run bootstrap, or no lifecycle action without executing lifecycle, token, import, or UI effects.
+- Added first-class policy explanations: `security action:"explain"` and lower-level `agent_harness mode:"policy_explain"` show why one model action is allowed, denied, or waiting on confirmation across the Agent route guard, permission mode, and typed tool confirmation, with secret-looking args redacted and no tool execution.
+- Added user-first delegation decision cards and structured handoff briefs: execution and delegation posture now distinguish local-first work, TUI handoff, delegated review, remote inspection, and blocked hidden fanout, while `/delegate` and the Agent Workspace form preserve reason, success criteria, workspace hints, priority, and explicit review intent.
+- Added the first-class `delegation` tool: `action:"status|routes|route"` exposes delegation policy, route catalogs, and exact confirmed handoff contracts without requiring harness mode names.
+- Added the first-class `execution` tool: `action:"status|route|history|record|processes|process_capabilities|process|recovery"` exposes local-vs-delegated work posture, exact route inspection, activity cards, tracked process inspection, direct process parity/doctor reporting, and file recovery without requiring harness mode names.
+- Added the first-class `computer` tool: `action:"status|plan|control|browser|setup|mcp|open_browser"` exposes browser/PWA readiness, browser/screenshot/desktop-control route planning, repair/setup rows, trusted MCP/tool discovery, and confirmed browser cockpit opens without requiring harness mode names.
+- Added normalized autonomy queue controls: research runs, connected-host tasks, approvals, automation runs, and schedules now expose available/unavailable controls with reasons plus exact confirmed routes for checkpoint, pause, resume, cancel, approve, deny, retry, run, edit, enable, disable, and delete actions; host task cancel/retry stays on confirmed `agent_operator_method` routes while `/tasks` remains inspection-only.
+- Added the first-class `autonomy` tool: `action:"intake|queue|item|status"` exposes ongoing-work route selection and visible autonomy queue inspection without requiring harness mode names.
+- Added trigger workflow posture: `autonomy action:"intake"` now maps time-based wakeups/schedules, incoming webhook/event watchers through published `watchers.*` daemon routes, Gmail/email connector-gated triggers, and read-only control-plane event streams, with watcher creation kept admin-confirmed and source-scoped.
+- Added certified watcher receipts: `autonomy action:"intake"` now exposes watcher success criteria, and `agent_operator_method` summarizes `watchers.create/patch/run/start/stop/delete` receipts into certified or follow-up outcomes without exposing operator tokens.
+- Added a source-owned watcher evidence contract: `autonomy action:"intake"` now tells users and models which SDK/daemon-owned durable run-history receipts, provider source records, redacted event payload descriptors, and queue correlation records are needed before Agent claims persisted webhook/Gmail watcher history.
+- Added host task output posture: autonomy queue connected-host task records now expose `/tasks output` routes plus bounded redacted result/error previews when the host publishes them, and route-only status when true output text is not in the read model.
+- Completed package and release hardening for the release gate: Bun-only install/run instructions, package-facing text verification, package runtime bundling, packed global install smoke, blocked lifecycle command smoke, source/package boundary checks, architecture checks, recorded performance snapshot checks, verification ledger, strict live verification artifacts, and release-quality readiness inventory dimensions are now part of the validated release path.
+- Completed the compact model-visible harness pass: `agent_harness` summary now starts with an assistant-first cockpit before technical counters, plural catalogs default to compact rows, `mode:"modes"` searches every harness mode by task/id/effect/parameter, and `mode:"mode"` inspects one mode contract with full schemas, policy blocks, route metadata, editor fields, release artifact detail, redacted log tail, and parameter detail behind `includeParameters:true` or singular inspect modes.
+- Added a first-class `workspace` adapter for workspace actions, panels, UI surfaces, slash commands, CLI mirrors, keybindings, and fixed shortcuts, so the model can inspect or confirm visible workspace operations without memorizing harness mode names; lower-level harness modes remain for detailed route inspection and compatibility.
+- Added ordered channel setup guide: Agent Workspace -> Channels, `/channels guide`, and `channels action:"setup"` now rank the next channel and walk through setup schema, secret-backed settings, delivery target, allowlist policy, live status/doctor checks, and explicit confirmed test-send boundaries.
+- Added channel triage: Agent Workspace -> Channels, `/channels triage`, and `channels action:"triage"` now unify setup blockers, daemon delivery attempts, visible surface messages, route bindings, and redacted Agent receipts without claiming provider-specific unread inbox polling.
+- Added channel delivery receipts: confirmed `/channels send` and `agent_channel_send` deliveries now write Agent-owned redacted receipt history, visible through Agent Workspace -> Channels, `/channels deliveries`, and `channels action:"deliveries"`.
+- Added route-backed browser and desktop control setup posture: `setup action:"status|item"` plus `execution action:"status|route"` now show ready, attention, or setup-needed state, workflow cards, setup checklists, fallback routes, and MCP review routes for browser, desktop, computer-use, screenshot, or screen-recording tooling; lower-level `setup_posture`, `setup_item`, `execution_posture`, and `execution_route` remain compatibility routes.
+- Added connected browser cockpit routing: Home, `computer action:"browser|open_browser"`, `workspace action:"surface|open"`, and lower-level UI surface detail routes now expose the configured connected-host browser cockpit/PWA URL with service/web setup fallback, workspace-category coverage, mobile/PWA controls, Agent onboarding marker status, and an honest unpublished browser/PWA first-run receipt contract.
+- Added companion device capability mapping: `device action:"status|capability"` and lower-level `pairing_posture` / `pairing_route pairingRouteId:"device-capability-map"` now show ready/attention/setup-needed/not-published posture for companion pairing, mobile command routing, browser/PWA, voice controls, spoken responses, notifications, browser/desktop control, and camera/location sensors without returning raw tokens or claiming unpublished device APIs.
+- Added voice workflow posture: `device action:"voice"` and lower-level `media_posture` now map push-to-talk input, voice memo transcription, spoken responses, and wake-word capture with ready/attention/setup-needed/not-published state, setup routes, runtime evidence, and explicit policy that always-listening capture is not published yet.
+- Surfaced voice/device/browser posture in Voice & Media: the workspace now has Voice workflows, Device capability map, and Browser/PWA readiness actions with direct `device` and `computer` route hints.
+- Added sudo execution posture: `execution action:"status|route"`, `setup action:"status"`, `setup action:"item" setupItemId:"sudo-execution-posture"`, and `process action:"capabilities"` now expose foreground-supervised escalation guidance, SUDO_PASSWORD presence-only status, blocked background sudo/stdin password routes, and missing SDK/daemon mediation contracts without reading or printing raw password values.
+- Added connected-host setup repair cards: expanded `setup_posture` and `setup_item` rows now expose status/service diagnostics plus confirmed service install/start/restart `agent_operator_method` routes when the SDK operator contract supports them, while first-run setup excludes service stop/uninstall.
+- Added probe-fed setup repair recommendations: connected-host setup now includes live service probe evidence, recommends diagnostics/status first, and keeps service install/start/restart as inspect-first confirmed routes until service status proves they are needed.
+- Added certified service repair receipts: setup repair cards now include success criteria and verification routes, and `agent_operator_method` summarizes services.status/install/start/restart receipts into certified or follow-up outcomes without exposing the operator token.
+- Added service lifecycle receipt decisions: setup posture now exposes serviceLifecycleDecision gates, and `agent_operator_method methodId:"services.status"` maps returned installed/running/control-plane evidence to exact install, start, restart, or no-action guidance before any lifecycle mutation.
+- Added connected-host auth setup posture: first-run setup now has a token-safe connected-host auth row for missing/usable operator token state, fingerprint-only token evidence, `/auth review` guidance, exact pairing route ids, confirmed SDK-backed local token create/repair, and regression fixtures for missing-host, reachable-host, missing-token, and unconfigured-model setup paths.
+- Added setup smoke execution: first-run setup now includes a token-safe install smoke row plus confirmed `setup action:"smoke"` evidence collection from package binary/version/status to connected-host status, auth posture, provider/model routing, setup posture, one first assistant turn, optional saved redacted evidence artifacts, and Home/setup latest-result plus history/trend/frequent-blocker surfacing without running shell commands implicitly.
+- Added guided setup wizard: the Start workspace and `setup_posture` now expose progress, current-step routes, backtracking routes, setup-smoke rerun/save routes, and repeated-blocker focus from saved smoke evidence so first-run repair feels like one guided flow instead of a loose checklist.
+- Added saved setup wizard checkpoints: Start and `setup action:"status"` now show the saved resume state, stale checkpoint auto-advance evidence, Start show/save/clear checkpoint actions, and `setup action:"checkpoint|save_checkpoint|clear_checkpoint"` inspect or mutate only Agent-owned step resume metadata after confirmation.
+- Added setup closeout: Start and `setup action:"status"` now expose `setupWizard.closeout` / `setupCloseout` decisions that reduce critical setup blockers, saved setup smoke evidence, and the user onboarding completion marker into blocked, needs-smoke-evidence, ready-to-finish, or complete states; confirmed `setup action:"finish"` writes only the user onboarding markers.
+- Added first-run local model readiness: `setup action:"status"` and `setup action:"item"` now include a local-model readiness row with detected local stacks/routes, top cookbook recipe, readiness score, benchmark follow-through, and exact model-routing inspect routes.
+- Added Personal Ops connector tool classification: expanded inbox and calendar lanes now read MCP tool metadata when available, classify read-only versus write-like email/calendar tools, and carry those capability tags into connector records and workflow prerequisites.
+- Added Personal Ops schema-derived operation records: expanded inbox and calendar lanes now turn reviewed MCP schemas into required-field/sample-input/schema-route cards with explicit confirmation flags for send/edit-like actions.
+- Added Personal Ops execution plans: request intake now returns ordered connector-read, local-compose, setup-repair, and confirmed-effect steps so inbox/calendar work is legible before any live provider action.
+- Added confirmed Personal Ops read execution: `personal_ops action:"read"` (backed by `run_personal_ops_read`) runs one selected read-only inbox/calendar MCP operation after required-field and confirmation checks, refuses write-like tools, and returns bounded redacted output plus normalized review cards; users can also request a saved redacted review-card artifact without storing full raw connector output or full input values.
+- Added next-route packets to confirmed Personal Ops reads so refreshed reads, lane/queue inspection, saved artifact reopen, local reply/reminder drafting, and send/calendar-edit boundaries are explicit structured routes after one read completes.
+- Added Personal Ops saved review queues: saved inbox/calendar review artifacts now resurface as redacted thread/event queue records with artifact inspect routes, freshness status, confirmed refresh routes when a matching read connector is ready, local draft/reminder follow-up routes, and explicit confirmed provider-effect boundaries.
+- Added the direct Personal Ops review queue: `personal_ops action:"queue"` aggregates saved inbox thread/calendar event review items, fresh provider-read routes, refresh routes, and follow-up confirmation boundaries without executing MCP tools or mutating artifacts.
+- Added Personal Ops task/reminder workflows: task lanes now separate visible work-plan actions from connected-host task inspection/control, and reminder lanes now expose confirmed reminder creation, autonomous schedule creation, schedule review, run, pause/resume, and exact-id control records.
+- Added Personal Ops daily briefings: `personal_ops action:"briefing"` and the visible Daily briefing plan action now assemble one read-only user-first plan across inbox, agenda, tasks, reminders, routines, delivery, notes, and autonomy queue before any live personal-data read or effect.
+- Added the first-class `personal_ops` adapter: `action:"briefing|status|queue|intake|lane"` exposes the safe Personal Ops discovery flow directly, while confirmed `action:"read"` preserves the existing one-operation inbox/calendar read boundary with required fields, redaction, and optional saved review cards.
+- Added research runner posture: `research_runs` now exposes browser-backed research readiness, setup and fallback routes, source review, report save, Knowledge promotion routes, and policy so deep research requests do not pretend browser execution is ready when only web/fetch research is available.
+- Added visual research report packets: `research action:"report" visualReport:true` now appends at-a-glance, evidence matrix, findings board, dated source/comparison, open-question, next-action, and handoff-checklist sections to the saved sourced artifact while preserving citation coverage metadata and no transcript body dump.
+- Added guided learning consolidation execution: `memory action:"curator"` now returns an ordered duplicate-consolidation batch plan and first-class `agent_learning_consolidation` preview/merge/stale/delete/rollback/recreate routes; merge and stale phases write durable receipts, delete refuses duplicates that have not been staged stale, post-delete receipts preserve snapshots with exact-id recreate guidance, and confirmed recreate refuses unsafe id collisions before restoring deleted duplicates.
+- Added learning prompt plans: `memory action:"curator"` now returns prompt-active records, suppressed review/setup/low-confidence/personality/consolidation counts, proposal queues, consolidation queues, and usefulness/freshness/source-quality/risk ordering rules so users and the model can see what may guide the assistant now before durable context expands.
+- Added prompt context inspection and receipts: prompt builds now write durable sanitized receipt ids with turn/source/model/provider, selected and suppressed record refs, segment counts, prompt hash, size, timestamp, and completed/error/cancelled outcome without storing raw prompt or response text; Agent Workspace -> Local Context shows a compact receipt outcome timeline with exact latest-receipt drill-in and outcome filter routes, and `context action:"prompt|receipts|receipt"` exposes recent receipts, exact `receiptId`/`turnId`/`outcomeStatus` filters, turn outcomes, applied prompt composition order, selected VIBE.md/project context/memory/routine/skill/persona records, suppressed records, prompt previews on request, and approximate token budget without mutating local behavior.
+- Added missing-host setup bootstrap: `setup action:"item"` for connected-host readiness now returns user-run GoodVibes host install, trust, binary verification, service start, and Agent reconnect commands before operator methods are reachable.
+- Added first-class model/provider routing: `models action:"status|route|local|providers|provider|smoke"` now fronts model route readiness, provider/subscription posture, local model cookbook, exact route/provider lookup, and confirmed local server smoke checks while lower-level harness modes remain available for detail.
+- Added model readiness scoring: `model_routing`, `model_route`, and the local model cookbook now expose estimated 0-100 readiness across latency, context window, tool support, vision, cost, and privacy while clearly flagging missing live benchmarks; the visible Model Routing lane now has an `Inspect route readiness` action for that read-only posture.
+- Added local server endpoint checks: the local model cookbook now maps detected local provider/model registry endpoints and local base URL environment hints into exact endpoint inspect routes, model-list smoke commands, success criteria, failure triage, confirmed `models action:"smoke"` checks for detected or default local endpoints, refresh routes, provider-add hints, suggested defaults, and a visible `Check local servers` workspace action without running a hidden network probe.
+- Added local model benchmark execution and evidence: the Model Routing workspace now includes a confirmed local benchmark action backed by `agent_model_compare` with `benchmarkKind:"local-model-route"` plus a `Review benchmark evidence` action for saved comparisons, revealed winner judgments, and filtered analytics before any separate default-model apply action.
+- Added the visible assistant cockpit to TUI Home: the existing renderer now opens on the same setup, chat/model, project work, Personal Ops, research/docs, background supervision, and safety/recovery lanes as `agent_harness mode:"summary"`, while deeper subsystem details stay in categories and search.
+- Tightened local behavior prompt quality gates: reviewed memory now requires the durable confidence threshold, and skills, routines, bundles, and personas only steer the assistant when reviewed and setup-ready; suppressed behavior remains visible as review/setup work.
+- Added memory posture UX: `memory action:"status|provider|list|search|get"` now exposes Agent-local memory counts, direct record lookup/search, prompt-active recall, vector stats, embedding-provider doctor warnings, provider inspection, and external-memory setup contract maps for Honcho, OpenViking, Mem0, Hindsight, Holographic, RetainDB, ByteRover, and Supermemory, including provider-specific next routes, missing setup/status/read/write/sync checklist items, and required sync/write receipt fields while honestly marking provider records as not published until the SDK/daemon exposes concrete setup/status/read/write/receipt records for Agent to consume.
+- Added background-process parity reporting: `process action:"capabilities"` now maps terminal background start plus process list/poll/wait/log/kill/write, PTY, and sudo semantics to supported routes or explicit SDK/daemon contract gaps, with lower-level lifecycle aliases still accepted for compatibility.
+- Improved background process logs: `execution action:"process"` and `process action:"log"` now return redacted stdout/stderr tails with byte counts, character counts, truncation flags, and direct process-style routes.
+- Added first-class terminal/process adapters: `terminal command:"..." background:true` starts a visible tracked local process, while `process action:"list|poll|wait|log|kill|write"` manages the same ProcessManager records with the existing confirmation and unsupported PTY/stdin/sudo boundaries.
+- Added a first-class schedule adapter: `schedule action:"list|create|remind|edit|run|pause|resume|delete"` routes recurring work, reminders, exact schedule edits, and lifecycle controls through the existing preview, confirmation, current-state diff, and allowlisted connected-host schedule paths.
+- Added post-action next routes for schedules: confirmed autonomous schedule creation, routine promotion, reminders, schedule edits, and schedule lifecycle controls now return schedule list, autonomy queue, run, edit, pause, resume, and delete routes without weakening confirmation gates.
+- Added pre-confirmation schedule handoffs: autonomous schedule, reminder, routine-promotion, and schedule-edit previews now include explicit confirmation routes while remaining non-mutating.
+- Added a first-class setup adapter: `setup action:"status|item|repair|checkpoint|save_checkpoint|clear_checkpoint|token|smoke|finish|import_settings"` gives first-run setup a short user-intent route while delegating to the existing wizard, repair decision, token, smoke, closeout, and settings-import implementations.
+- Added a first-class VIBE.md adapter: `vibe action:"status|show|init|import_persona"` gives personality setup a short user-intent route while preserving secret scans, confirmation gates, and Agent-local persona boundaries; unconfirmed init/import previews now include exact model and CLI `confirmationRoutes`.
+- Added a first-class settings adapter: `settings action:"list|get|set|reset|import"` fronts Agent settings and GoodVibes settings import preview/apply without requiring the model to know harness settings modes or the workspace action id.
+- Shared GoodVibes settings import previews now include source-package ownership metadata so users can reuse goodvibes-tui or other published GoodVibes platform settings without Agent mutating source package stores.
+- Added a first-class host adapter: `host action:"status|capabilities|capability|services|service|methods|method"` fronts connected-host status, capability maps, service posture, endpoint detail, and daemon method inspection while preserving exact execution on `agent_operator_method`.
+- Added a first-class device adapter: `device action:"status|capability|browser|control|voice|provider|open_browser|open_tts_provider|open_tts_voice"` keeps companion, mobile/PWA compatibility, voice/TTS, browser/desktop-control compatibility, provider posture, and visible browser/TTS picker handoffs available while preserving confirmation gates; `computer` is now the primary browser/PWA and desktop-control route.
+- Added a first-class research adapter: `research action:"briefing|plan|search|runner|runs|run|sources|source|bundle|reports|report_artifact|create_run|start_run|checkpoint|pause|resume|cancel|complete|fail|delete_run|add_source|review_source|reject_source|use_source|delete_source|report"` gives deep research one user-facing route over a read-only next-action queue, workflow planning, bounded public source-candidate search, browser-runner readiness, visible run lifecycle, source triage, reviewed-source bundles, saved report inspection, and sourced report artifacts while preserving existing confirmation gates.
+- Added dynamic background-process substrate probing: the process parity report now detects ProcessManager stdin/PTY methods, terminal/PTY operator methods, GoodVibes session-input routes, and credential routes, and `processAction:"write"` dispatches through a discovered safe ProcessManager stdin method only after explicit confirmation without echoing input data.
+- Added Work workspace process supervision: Work & Approvals now shows tracked/running/completed local process counts, stdin/PTY/sudo parity, process monitor/live-tail routes, and visible Background processes / Process capabilities actions.
+- Added deep research briefing: `research action:"briefing"` now returns one read-only next-action queue across visible runs, source review, saved report artifacts, browser readiness, and exact follow-up routes without searching, mutating, opening browser surfaces, saving reports, ingesting Knowledge, exporting artifacts, archiving, sharing, or sending messages.
+- Improved run-bound source search: `research action:"search" runId:"..."` now uses the visible run's saved question and returns run-specific inspect, briefing, start, and checkpoint follow-up routes without creating, starting, checkpointing, or mutating source records.
+- Added cross-project handoff docs for GoodVibes TUI/daemon and GoodVibes SDK so browser-backed research, browser/PWA rendering, Personal Ops provider queues, PTY/sudo/process, watcher history, remote runner evidence, local model health, mobile/voice/device, channel receipts, and external memory provider contracts have concrete acceptance criteria before Agent consumes them.
+- Added next-route packets to research run detail and mutation outputs so create/start/checkpoint/pause/resume/cancel/complete/fail/delete results point to the next visible inspect, briefing, workflow, run-bound search, source queue, checkpoint, report save, artifact inspection, or Knowledge promotion route without performing hidden work.
+- Added next-route packets to research source detail and mutation outputs so add/review/reject/use/delete results point to the next visible inspect, queue, review/reject, bundle, sourced report save, mark-used, report artifact, or optional Knowledge promotion route while preserving separate confirmation gates.
+- Added next-route packets to confirmed research report saves so saved artifacts point to high-level report inspection, artifact export/archive, Knowledge promotion, report listing, and visible run completion when `runId` is provided.
+- Added deep research workflow planning: `research action:"plan"` now returns a read-only ordered route plan across visible run state, public web/fetch or browser posture, source capture/review, sourced report saves, and optional artifact-to-Knowledge promotion.
+- Added research runner/report contracts: `research action:"runner"` directly exposes browser-runner setup and user-control requirements, while `research action:"plan"` includes those contracts plus visual-report packet sections, citation/source-map acceptance criteria, and artifact archive routes so missing browser execution and browser/PWA report rendering are explicit product gaps rather than hidden assumptions.
+- Added Research workspace contract visibility: the existing renderer now shows browser-runner and visual-report readiness in the Research area, keeps the direct research run/source/report routes visible, and adds Research briefing, Plan workflow, Public source search, Browser runner readiness, and Report artifacts actions that map directly to `research action:"briefing"`, `research action:"plan"`, `research action:"search"`, `research action:"runner"`, and `research action:"reports"`.
+- Added artifact ZIP archive export: `agent_artifacts mode:"archive"` writes the same reviewed package payload as one workspace ZIP file, while the existing package workspace form now lets users choose directory or ZIP output with the same confirmation and redacted manifest policy.
+- Added reviewer-ready document export appendices: `agent_documents mode:"export"` now saves comment and AI suggestion summaries plus review metadata counts with the markdown artifact, without printing draft content in the tool response.
+- Added reviewer-readiness checks: Document Ops now flags missing source artifacts, unresolved comments, proposed suggestions, unrevealed comparisons, hidden judgments, route-change decisions, and incomplete handoff evidence before export, archive, or model-route apply.
+- Added visible reviewer preflight and handoff diff forms: Documents & Compare now has a read-only readiness preflight before export/archive/apply and a split-pane reviewer handoff diff form backed by `agent_model_compare mode:"handoffDiff"` with section jumps for metadata, policy, related artifacts, or comparison evidence.
+- Added recent handoff choices in the diff form: Documents & Compare now reads saved reviewer handoff artifact metadata, shows the saved count, pre-fills the newest older-to-newer pair when available, and still lets blank IDs list recent handoffs safely.
+- Added inline reviewer-readiness badges: document export, reviewer handoff/archive, and apply-winner forms now show readiness status, issue counts, and the next preflight action at the point of confirmation.
+- Added review packet timeline: Documents & Compare and `document_ops_lane` now show one chronological document/comment/suggestion/attachment/export/compare/judgment/handoff/archive packet history with the next route decision.
+- Added guided review packet defaults: Documents & Compare now picks the latest packet document/export/comparison/judgment/handoff evidence and pre-fills document export, compare handoff/archive, and apply-winner forms with editable field hints while preserving confirmation gates.
+- Added reusable review packet presets: Documents & Compare now has a confirmed Save packet preset form backed by `agent_review_packet_presets`, with local JSON preset artifacts in the packet timeline, list/show reuse routes, and default fallback only when live packet evidence is missing.
+- Added review packet preset freshness checks: preset list/show routes and the packet timeline now flag missing or superseded saved artifact ids, and show recommended reuse routes with newer matching evidence when metadata is sufficient.
+- Added review packet preset refresh: stale presets can now be refreshed through a confirmed workspace form or `agent_review_packet_presets mode:"refresh"`, saving a new local preset artifact with source-preset lineage instead of making the user copy replacement ids by hand.
+- Added review packet wizard: Documents & Compare and `document_ops_lane` now expose a read-only six-step packet guide across draft review, document export, compare judgment, reviewer handoff, route decision, and final archive review with progress, current-step route hints, backtracking routes, and refreshed-preset lineage before sharing.
+- Added review packet sharing: Documents & Compare now has a confirmed Share review packet form backed by `agent_review_packet_share`, which validates a saved reviewer handoff archive, previews the configured delivery target and packet evidence ids, and sends only a plain-text archive reference after explicit confirmation.
+- Added route-decision receipts: confirmed `agent_model_compare mode:"apply"` now saves an apply-winner receipt artifact, `mode:"routeDecision"` records leave-unchanged decisions without changing model routing, and Document Ops readiness/wizard state clears matching route-decision blockers before archive evidence exists.
+- Added route-decision receipt archive evidence: `agent_model_compare mode:"handoffArchive"` now auto-includes matching route-decision receipt artifacts in reviewer handoff ZIP archives, README counts, archive metadata, and redacted manifests.
+- Added blind comparison synthesis: `agent_model_compare mode:"synthesis"` groups saved judgment reasons into stable cross-session preference themes, reports revealed/hidden winner posture, and routes through the existing Compare Analytics workspace form without changing model routing.
+- Added blind comparison reviewer handoffs: `agent_model_compare mode:"handoff"` creates one confirmed markdown artifact that combines saved comparison or judgment evidence with related document/artifact exports, while the existing Export Compare workspace form can choose report or handoff.
+- Added blind comparison handoff archives: `agent_model_compare mode:"handoffArchive"` turns a saved reviewer handoff into one ZIP artifact with the handoff, source comparison or judgment, related evidence bytes, README, and redacted manifest.
+- Added side-by-side comparison review: `agent_model_compare mode:"sideBySide"` renders related document/artifact excerpts beside saved comparison or judgment evidence through the existing Review Saved Compare form without creating artifacts or changing routes.
+- Added filtered comparison analytics: saved comparison judgments now preserve task type, document id, and benchmark tags so `agent_model_compare mode:"analytics"` and `mode:"synthesis"` can show targeted preference trends without changing routes.
+- Added VIBE.md personality support: project/global VIBE.md files are discovered, secret-scanned, injected into the serial Agent prompt, surfaced in the Personas workspace, and manageable through `vibe action:"status|show|init|import_persona"` or `/vibe status|init|show|import-persona`.
+- Added project context support: `.hermes.md`, `HERMES.md`, `AGENTS.md`, `CLAUDE.md`, optional `HERMES_HOME/SOUL.md`, `.cursorrules`, and `.cursor/rules/*.mdc` are secret-scanned, injected as project instructions separate from VIBE.md personality, and exposed through `context action:"files|file"` with target-aware subdirectory lookup.
+- Surfaced VIBE.md and project context health in the Local Context and Personas workspace: users now see applied/blocked/truncated counts and direct Inspect VIBE.md, Inspect project context, and Inspect one context file actions without knowing harness modes.
+- Added managed Agent orchestration cards: `agent_harness mode:"agent_orchestration"` and `mode:"agent_orchestration_agent"` expose live subagent state, serial-by-default policy, multi-agent milestones, per-agent plan cards, remote-runner contract/artifact evidence, spawn/batch-spawn decision cards, templates, and exact first-class `agent` inspect/message/wait/cancel routes without creating hidden work.
+- Added confirmed work-plan agent dispatch: `agent_work_plan action:"dispatch_agents"` previews selected approved work-plan items, then calls first-class `agent` spawn or batch-spawn after confirmation and writes linked-agent receipts back to the plan.
+- Added post-dispatch next routes for work-plan agent dispatch: successful dispatch output and linked work-plan detail now point directly to orchestration, work-plan, agent inspect/wait/message/cancel, and closeout routes.
+- Added orchestration closeout evidence: managed Agent cards now correlate linked work-plan items, dispatch receipt counts, closeout update routes, and remote-runner artifacts auto-attached by runner id with exact review routes.
+- Added VIBE.md setup and curator health: blocked or truncated project/global personality files now appear in first-run setup and `memory action:"curator|candidate"` cards with `/vibe` inspection, init, and import routes.
+- Added VIBE.md profile starter portability: starter export/from-discovered flows can include safe VIBE.md content with `--include-vibe`, imports preserve it, and profile creation writes it into the isolated profile home.
+- Added tracked local background process controls: `execution action:"processes|process"`, `terminal command:"..." background:true`, and `process action:"poll|log|wait|kill|write"` expose shared ProcessManager start/list/status/log/wait/stop with bounded redacted output, process monitor/live-tail handoffs, session-id aliases, conditional confirmed stdin-write dispatch when a safe ProcessManager method exists, unsupported PTY guidance, and blocked hidden sudo prompts; lower-level background harness modes remain compatibility routes.
+- Added execution history activity cards: `execution action:"history"` groups redacted tool/shell/edit records into user-facing outcome cards with verification evidence, bounded process-output summaries, supervision routes, and file-recovery handoffs, while `execution action:"record"` keeps exact record inspection available.
+- Completed connected-host daemon aliases for the model: `daemon` maps to connected-host posture and `daemon_status` maps to live connected-host status while lifecycle control stays outside Agent.
+- Completed settings parity for the model-facing harness and then promoted it behind first-class `settings`: discovery is compact by default with per-setting `modelRoute` hints, `settings action:"get"` and `includeParameters:true` expose full descriptors, and `settings action:"set|reset"` use the same config/secret managers with confirmation and external-host setting locks.
+- Completed the tool-description verbosity sweep: first-class Agent tools and wrapped built-in tool definitions now register concise descriptions, nested parameter schema descriptions are stripped from the default model-visible catalog, and detailed safety rules remain enforced in policy wrappers, confirmation errors, and detailed harness inspection.
+- Completed copyable model route hint cleanup: confirmed harness examples now include the required `confirm:true` and `explicitUserRequest` fields so route hints match the actual execution contract.
+- Completed resilient shortcut/keybinding discovery: `agent_harness` returns fixed shortcuts and default-fallback keybinding descriptors when the live keybinding manager is absent, while keybinding execution and mutation still fail closed until the runtime manager is available.
+- Completed the model-visible operator method catalog: `agent_harness` can now list and inspect allowlisted public operator and Agent Knowledge methods with their owning first-class model tools, confirmation policy, and boundaries, without exposing arbitrary route invocation.
+- Expanded first-class Agent Knowledge reads: `agent_knowledge` now covers status, ask, search, source/node/issue lists, item lookup, map summary, connector list/detail, and connector doctor while staying inside `/api/goodvibes-agent/knowledge/*`.
+- Added Agent-owned Knowledge response-scope validation: CLI, model tools, and live verification normalize public Agent-route scope aliases and reject successful-looking payloads with known non-Agent payload markers.
+- Expanded release-route enforcement for Agent Knowledge: architecture/package policy now guards the full route catalog, and live verification definitions include source, node, issue, map, and connector read routes in addition to status, ask, and search.
+- Verified the release line against connected-host public Agent routes for Agent Knowledge alias scope normalization and telephony channel surface support.
+- Renamed active release evidence artifacts to current-release paths under `release/`, including release notes, readiness inventory, performance snapshot, and live-verification reports, so current release metadata no longer depends on stale historical filenames.
+
+## 1.1.7 - 2026-06-05
+
+- Hardened the Agent model tool surface so goodvibes_context routes to the Agent harness instead of dead-ending.
+- Added tool execution and permission safety guards so registered model tools return structured failures instead of aborting turns.
+- Added `route action:"plan|status"` and `agent_harness mode:"route_decision"` for read-only user-task route planning across visible Agent surfaces.
+- Routed host/daemon health, doctor, readiness, service, and compatibility wording through `host action:"status"` before setup repair or service lifecycle effects.
+- Routed normal settings/configuration wording through `settings action:"list"` so set/reset/import requests start with Agent-owned setting discovery and explicit confirmation boundaries.
+- Routed model provider, local-cookbook, local server smoke, and route-fit wording through `models action:"provider|local|smoke|route"` before credential, smoke, benchmark, or route-change effects.
+- Routed direct reminder, schedule, cron, and schedule lifecycle wording through `schedule action:"list"` plus confirmed schedule actions, while broader ongoing work remains on autonomy intake.
+- Added a source-owned watcher evidence contract to autonomy intake so durable run-history receipts, provider source records, redacted event payload descriptors, and queue correlation stay explicit SDK/daemon handoff requirements instead of hidden Agent assumptions.
+- Routed Personal Ops briefing, saved queue, fresh inbox/calendar read, and connector setup wording through `personal_ops action:"briefing|queue|intake|lane"` before live provider reads or effects.
+- Added `setup action:"repair"` and `agent_harness mode:"setup_repair"` for read-only setup repair decisions that choose token repair, host status, services.status receipt, user-run bootstrap, or no lifecycle action without executing effects.
+- Added `execution action:"capabilities|process_capabilities"` so process parity, PTY, stdin, and sudo posture reports are reachable from the first-class execution tool.
+- Added `computer action:"plan"` and `agent_harness mode:"browser_control_route"` for read-only browser navigation, screenshot, and desktop-control workflow routing; `route action:"plan"` now prefers that computer planner for screenshot, browser-control, screen-observation, and desktop-control wording.
+- Routed command-shaped background work through local process supervision so `route action:"plan"` prefers `execution action:"processes"` plus first-class `terminal`/`process` controls instead of delegation or autonomy for local long-running commands.
+- Routed interactive terminal, PTY, stdin, and sudo wording through `execution action:"process_capabilities"` so unsupported interactive routes are explained before any hidden process start or credential effect.
+- Routed external memory-provider, backend, cross-session sync, import/export, and named-provider wording through `memory action:"provider"` or the external provider checklist before promising provider writes, sync, credentials, or import/export effects.
+- Added provider-specific next routes, missing setup/status/read/write/sync checklist items, and required receipt fields to the external memory-provider posture so Agent can explain exactly what SDK/daemon records are missing before provider-backed memory is used.
+- Clarified shared GoodVibes settings import with source-package ownership metadata so imports remain user-controlled without implying Agent owns the source package capability.
+- Routed browser-backed research runner wording through `research action:"runner"` and visual research report rendering wording through `research action:"plan"` plus report artifacts before claiming browser/PWA rendering readiness.
+- Routed voice workflow, TTS-provider, browser cockpit, and PWA wording through `device action:"voice|provider"` or `computer action:"browser"` before capture, playback, picker, or browser-open effects.
+- Routed channel setup, triage, delivery receipts, and send wording through `channels action:"setup|triage|deliveries|channel"` before confirmed external delivery.
+- Routed file undo/redo/recovery wording through `execution action:"recovery"` so plain repair requests inspect FileUndoManager snapshots before any confirmed undo or redo is applied.
+- Routed media generation wording through provider readiness and confirmed `agent_media_generate` saved-artifact output so generated bytes stay out of transcripts and Knowledge unless explicitly promoted.
+- Routed permission posture, security findings, and blocked-action questions through `security action:"status|finding|explain"` so users can ask what is active, what happened, or why something was blocked without knowing policy tool names.
+- Added direct read-only `support`, `sessions`, and `audit` adapters, and routed support-bundle, saved-session/bookmark, and release/audit evidence wording through `support action:"status|bundle"`, `sessions action:"list|get"`, and `audit action:"readiness|evidence|item|artifact"` before export/import/share, session lifecycle, or audit drill-in effects.
+- Added `security action:"explain"` and `agent_harness mode:"policy_explain"` for read-only allowed/denied/confirmation-required policy explanations.
+- Added registered-tool smoke coverage for the Agent-guarded platform tool roster.
+
 ## 1.1.6 - 2026-06-05
 
 - Fixed Import GoodVibes settings so it also imports active and pending provider subscriptions from the GoodVibes TUI user store into Agent-owned subscription state.

package/README.md

@@ -3,9 +3,9 @@
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 [![Status: 1.0.x](https://img.shields.io/badge/status-1.0.x-green.svg)](#install)
 
-GoodVibes Agent is the installable personal operator assistant TUI for GoodVibes. It is built for day-to-day operator work: chat, setup, profiles, routines, skills, personas, Agent-local memory and notes, isolated Agent Knowledge, channel readiness, voice/media setup, work-plan tracking, approvals, automation visibility, and explicit build delegation.
+GoodVibes Agent is the installable autonomous operator assistant for GoodVibes. It keeps the existing terminal renderer and workspace bones, but the product goal is different from a vibecoding harness: the user should experience one assistant that can chat, plan, remember, research, schedule, send, generate, run visible agents, and operate the GoodVibes daemon contract with clear confirmation gates.
 
-Agent connects to a GoodVibes host owned outside this package. It does not install, start, stop, restart, expose, or own that host.
+The GoodVibes daemon is Agent's capability runtime. It provides the operator API, schedules, channels, knowledge, media, remote execution, service posture, and long-running automation routes. Those capabilities stay in their owning packages; Agent consumes published SDK/daemon/TUI contracts and presents the user-first harness, route planning, confirmation gates, and receipts. Agent keeps shared GoodVibes settings import so users can reuse provider selections, UI preferences, permissions, subscriptions, surfaces, tools, and daemon endpoint settings from goodvibes-tui and other published GoodVibes platform stores instead of rebuilding setup by hand.
 
 ## Install
 
@@ -51,60 +51,75 @@ Workspace areas:
 - Home: normal assistant chat, operator briefing, model selection, setup, and health.
 - Conversation: context usage, inline `@file`/`@folder`/`@url` references, compaction, title/session save/load/search/export, bookmarks, paste/image/TTS helpers, undo/redo/retry, clear/reset, shortcuts, and keybindings.
 - Research: read-only web research, URL inspection, source triage, and explicit handoff into Agent Knowledge.
+- Documents & Compare: versioned document drafting, uploads, exports, source checks, generated media artifacts, artifact browsing/export, artifact reuse, and confirmed blind model comparison with delayed reveal, durable JSON comparison artifacts, saved review boards, side-by-side reviewer views, saved judgment artifacts, saved preference analytics/synthesis, markdown report export, reviewer handoff artifacts, one-click reviewer handoff ZIP archives, and confirmed winner route updates.
 - Artifacts: image attachment, conversation/session export, source-file ingest, source lookup, bookmark/browser-history import, and generated media artifacts.
+- Personal Ops: request intake for inbox, agenda, tasks, reminders, notes, routines, and delivery; inbox/calendar connector readiness; saved review queues; notes, work plans, host tasks, task/reminder workflow cards, connected schedule controls, routines, schedules, and delivery readiness in one daily operations area.
 - Setup: provider/model, compatibility, Agent Knowledge readiness, profiles, support bundles, subscriptions, and auth review.
 - Tools & MCP: MCP server setup, tool inventory, trust review, secrets, and settings.
 - Knowledge: isolated Agent Knowledge status, ask/search, source/node/issue libraries, item lookup, map review, connectors, ingest, review queue, and reindex.
-- Memory & Skills: local memory, scratchpad notes, learned behavior capture, personas, skills, routines, and schedule promotion.
+- Memory & Skills: VIBE.md personality, project context files, local memory posture, prompt-active recall, vector/embedding health, scratchpad notes, learned behavior capture, personas, skills, routines, and schedule promotion.
 - Channels: companion pairing, channel readiness, confirmed channel delivery, and confirmed webhook notification management.
 - Voice & Media: voice review, spoken response setup, image input, confirmed image/video generation, browser-tool posture, and provider readiness.
-- Automation: reminders, schedule status, routine promotion receipts, reconciliation, and exact confirmed approval/automation/schedule actions.
-- Build Delegation: explicit handoff to GoodVibes TUI for build, fix, implementation, or review work.
+- Automation: reminders, schedules, visible autonomous agents, routine promotion receipts, reconciliation, and exact confirmed approval/automation/schedule actions.
+- Operator Runtime: full GoodVibes daemon method discovery, read-only status routes, confirmed write/admin routes, and service posture.
 
 Press `/` inside the workspace to search actions by name, category, command, or detail. Slash commands and CLI subcommands remain power-user/scriptable mirrors; the workspace is the first-class user path.
 
 ## Model-Visible Harness
 
-The main Agent model can inspect and operate the Agent-controlled harness through Agent-owned tools. The important entrypoint is `agent_harness`.
+The main Agent model can inspect and operate the Agent-controlled harness through Agent-owned tools. The important first-run entrypoint is `setup`; the detailed catalog entrypoint is `agent_harness`.
 
-`agent_harness mode:"summary"` is compact by default. It returns counts, status, and a short mode guide. `mode:"modes"` searches every harness mode by task, family, effect type, id, alias, or parameter name; `mode:"mode"` inspects one mode contract. Plural catalog modes are also compact by default: they return ids, labels, counts, safe state, effect class, and short route hints. `workspace_actions`, `commands`, `cli_commands`, `panels`, `ui_surfaces`, `shortcuts`, `keybindings`, `tools`, `settings`, connected-host posture/status/capability, and operator/audit rows include compact `modelRoute` or `modelAccess` hints so the model can choose the right first-class tool or confirmed harness route without expanding every row. Use `includeParameters:true` or a singular inspect mode when the model needs schemas, detailed route hints, full policy blocks, redacted log tail, operator/audit artifact data, or editor field definitions.
+`route action:"plan" query:"..."` is the first stop when a plain user task could map to several GoodVibes surfaces. It returns the preferred visible route, alternatives, missing fields, confirmation boundary, workspace matches, and harness mode matches without running tools; host/daemon health or doctor wording routes to `host action:"status"`, normal settings/configuration wording routes to `settings action:"list"`, model provider/local-cookbook/smoke/route-fit wording routes to `models action:"provider|local|smoke|route"`, Personal Ops briefing/queue/fresh-read/connector wording routes to `personal_ops action:"briefing|queue|intake|lane"`, direct reminders/schedule lifecycle wording routes to `schedule action:"list"` and confirmed schedule actions, command-shaped background work routes to `execution action:"processes"` and the first-class `terminal`/`process` UX, interactive PTY/stdin/sudo wording routes to `execution action:"process_capabilities"` before any start or credential effect, external memory-provider/backend/sync/import/export wording routes to `memory action:"provider"` or the provider contract checklist, browser-backed research runner wording routes to `research action:"runner"`, visual research report rendering routes to `research action:"plan"` plus report artifacts, voice workflow and TTS-provider wording routes to `device action:"voice|provider"`, browser cockpit/PWA wording routes to `computer action:"browser"` before any visible open handoff, channel setup/triage/delivery-receipt/send wording routes to `channels action:"setup|triage|deliveries|channel"` before confirmed external delivery, security permission/status/finding/blocked-action wording routes to `security action:"status|finding|explain"` before policy changes or risky work, support-bundle wording routes to `support action:"status|bundle"` before bundle export/import/share effects, saved-session/bookmark/continuity wording routes to `sessions action:"list|get"` before session lifecycle effects, release readiness/evidence/audit wording routes to `audit action:"readiness|evidence|item|artifact"`, file undo/redo/recovery wording routes to `execution action:"recovery"`, media generation wording routes to provider readiness plus confirmed `agent_media_generate` artifacts, and screenshot, browser-navigation/control, screen-observation, and desktop-control wording routes to `computer action:"plan"` before any live UI tool is considered. `setup action:"repair"` is the first stop for setup or host fix requests; it chooses the safest next token repair, connected-host status, services.status receipt, user-run bootstrap, or no-action route without executing it. `agent_harness mode:"summary"` is compact by default. It returns counts, status, and a short mode guide. `mode:"modes"` searches every harness mode by task, family, effect type, id, alias, or parameter name; `mode:"mode"` inspects one mode contract, and `mode:"route_decision"` remains the lower-level compatibility route for the same user-task planner. Plural catalog modes are also compact by default: they return ids, labels, counts, safe state, effect class, and short route hints. Workspace, command, UI, panel, shortcut, and keybinding discovery should normally use `workspace action:"status|actions|action|run|surfaces|surface|open|commands|command|run_command|shortcuts|keybindings|keybinding"`, while lower-level harness modes remain for compatibility/detail inspection. Connected-host posture/status/capability and operator/audit rows include compact `modelRoute` or `modelAccess` hints so the model can choose the right first-class tool or confirmed harness route without expanding every row. Use `settings action:"list|get|set|reset|import"` for normal configuration work and `host action:"status|capabilities|capability|services|service|methods|method"` for normal connected-host/daemon diagnostics. Use `includeParameters:true` or a singular inspect mode when the model needs schemas, detailed route hints, full policy blocks, redacted log tail, operator/audit artifact data, or editor field definitions.
 
 High-value `agent_harness` mode groups:
 
-- Discovery: `modes`, `workspace_categories`, `workspace_actions`, `commands`, `cli_commands`, `panels`, `ui_surfaces`, `shortcuts`, `keybindings`, `tools`, `settings`.
-- Single-item inspection: `mode`, `workspace_action`, `command`, `cli_command`, `panel`, `ui_surface`, `keybinding`, `tool`, `get_setting`.
-- User-visible effects: `run_workspace_action`, `run_command`, `open_panel`, `open_ui_surface`, `run_keybinding`, `set_keybinding`, `reset_keybinding`, `set_setting`, `reset_setting`.
-- Product posture: `channels`, `notifications`, `provider_accounts`, `mcp_servers`, `setup_posture`, `model_routing`, `pairing_posture`, `delegation_posture`, `security_posture`, `support_bundles`, `media_posture`, `sessions`.
-- Connected host: `service_posture`, `service_endpoint`, `connected_host`, `connected_host_status`, `connected_host_capability`; `daemon` and `daemon_status` are aliases for connected-host posture/status.
-- Operator/audit inspection: `release_evidence`, `release_evidence_artifact`, `release_readiness`, `release_readiness_item`.
-- Operator methods: `operator_methods`, `operator_method`.
+- Discovery: `modes` and `tools`; use `workspace action:"status|actions|commands|cli_commands|panels|surfaces|shortcuts|keybindings"` for workspace/UI/command/key discovery and `settings action:"list"` for settings.
+- Single-item inspection: `mode` and `tool`; use `workspace action:"action|command|cli_command|panel|surface|keybinding"` and `settings action:"get"` for one item.
+- User-visible effects: use `workspace action:"run|run_command|open_panel|open|run_keybinding|set_keybinding|reset_keybinding"`, `terminal command:"..." background:true`, `process action:"wait|kill|write"`, and `settings action:"set|reset|import"`; all visible/mutating routes remain confirmation-gated.
+- Product posture: `channels action:"status|channel|setup|triage|deliveries"`, `notifications`, `provider_accounts`, `mcp_servers`, `setup action:"status|repair"` or detailed `setup_posture`, `host action:"status|capabilities|services|methods"`, `context action:"status|files|file|prompt|receipts|receipt"`, `memory action:"status|provider|curator|candidate|list|search|get"`, `agent_orchestration`, `model_routing`, `execution action:"status|route|history|record|processes|process_capabilities|process|recovery"`, `computer action:"status|plan|control|browser|setup|mcp"`, `personal_ops action:"briefing|status|queue|intake|lane|read"`, `document_ops`, `pairing_posture`, `delegation action:"status|routes|route"`, `security action:"status|finding|explain"`, `support action:"status|bundle"`, `media_posture`, `sessions action:"list|get"`.
+- Local long-running commands: `execution action:"processes"` and `execution action:"process"` inspect tracked ProcessManager jobs with bounded redacted output; `execution action:"capabilities"` or `action:"process_capabilities"` reports the same read-only process parity matrix as `process action:"capabilities"` for terminal background start, process list/poll/wait/log/kill/write, PTY, and sudo without overclaiming unsupported interactive routes. Confirmed `terminal command:"..." background:true` starts one visible local process, and `process action:"poll|log|wait|kill|write"` manages it. Lower-level `background_processes`, `background_process`, and confirmed `run_background_process` remain compatibility routes. `setup action:"item" setupItemId:"sudo-execution-posture"` shows SUDO_PASSWORD presence only, blocked background sudo routes, and foreground-supervised escalation guidance without reading or printing raw password values.
+- Visible Agent orchestration: `agent_orchestration` lists the live Agent manager, serial-by-default policy, managed multi-agent plan cards with milestones, work-plan links, dispatch receipts, closeout review routes, remote-runner contracts/artifact trails, auto-attached remote artifact review routes, spawn/batch-spawn decision cards, and safe control routes; `agent_orchestration_agent` inspects one visible subagent with its plan card. Approved work-plan items dispatch through confirmed `agent_work_plan action:"dispatch_agents"` into first-class `agent` spawn or batch-spawn calls with saved receipts. Actual spawn, message, wait, and cancel actions stay on the first-class `agent` tool.
+- Connected host: prefer `setup action:"repair"` for setup/repair decisions and `host action:"status|capabilities|capability|services|service|methods|method"` for diagnostics; lower-level `setup_repair`, `service_posture`, `service_endpoint`, `connected_host`, `connected_host_status`, `connected_host_capability`, `daemon`, and `daemon_status` remain compatibility/detail routes. The connected browser cockpit/PWA is surfaced through `computer action:"browser|open_browser"` and `workspace action:"surface|open"` with confirmation for visible opens; `computer action:"plan"` picks the safest browser, screenshot, or desktop-control workflow before live control tools are considered.
+- Project and prompt context: prefer `context action:"status|files|file|prompt|receipts|receipt"` for secret-scanned `.hermes.md`, `HERMES.md`, `AGENTS.md`, `CLAUDE.md`, `HERMES_HOME/SOUL.md`, `.cursorrules`, `.cursor/rules/*.mdc`, applied prompt composition, and receipt history; lower-level `project_context`, `project_context_file`, and `prompt_context` harness modes remain compatibility/detail routes.
+- Operator/audit inspection: `audit action:"readiness|item|evidence|artifact"`; lower-level `release_evidence`, `release_evidence_artifact`, `release_readiness`, and `release_readiness_item` remain compatibility/detail modes.
+- Operator methods: prefer `host action:"methods|method"` for inspection; exact execution stays on `agent_operator_method`.
 
 Every mutating or externally visible effect requires `confirm:true` plus `explicitUserRequest` unless a narrower first-class tool has its own confirmation contract. Ambiguous lookups return candidates instead of guessing.
 
 First-class model tools cover common workflows directly:
 
+- `route action:"plan|status"` for choosing the best visible Agent route for a plain user task before asking the user to understand Agent, TUI, host, daemon, or SDK ownership.
 - `agent_knowledge` and `agent_knowledge_ingest` for isolated Agent Knowledge reads and confirmed ingest.
-- `agent_local_registry` for Agent-local memory, notes, personas, skills, bundles, and routines.
-- `agent_work_plan` for visible local work-plan tracking.
+- `memory action:"status|provider|curator|candidate|list|search|get"` for Agent-local memory posture, provider readiness, review queues, and memory records; external-memory providers expose provider-specific next routes, missing setup/status/read/write/sync checklist items, and required receipt fields before Agent claims Honcho, Mem0, Supermemory, or similar backends are connected. `agent_local_registry` and `agent_learning_consolidation` remain the detailed routes for notes, personas, skills, bundles, routines, and confirmed duplicate-consolidation phases. Use `memory action:"status"` first when the user asks what memory knows, whether recall is healthy, or whether an external memory provider is actually connected.
+- `channels` for channel readiness, setup guide, triage, and delivery receipts; `agent_channel_send` remains the confirmed send route.
+- `support`, `sessions`, and `audit` for read-only support-bundle routes, saved-session/bookmark posture, release readiness, and packaged release evidence without requiring harness mode names.
+- `workspace` for Agent workspace categories/actions, slash and CLI command discovery, visible UI/panel navigation, shortcuts, keybindings, and confirmed workspace/command/keybinding effects.
+- `agent_work_plan` for visible local work-plan tracking and confirmed dispatch of approved plan items to visible agents.
+- `personal_ops` for daily briefing, request intake, lane inspection, and one confirmed read-only inbox/calendar connector operation.
+- `autonomy action:"intake|queue|item|status"` for ongoing-work route selection and visible autonomous-work queue inspection without requiring harness mode names.
 - `agent_operator_briefing` and `agent_operator_action` for connected work/approval/automation/schedule posture and exact confirmed actions.
-- `agent_channel_send`, `agent_notify`, `agent_reminder_schedule`, and `agent_media_generate` for confirmed delivery, notification, reminder, and media generation.
+- `agent_operator_method` for exact GoodVibes daemon contract parity. Read-only routes can run directly; write/admin routes require `confirm:true` and `explicitUserRequest`.
+- `agent_channel_send`, `agent_notify`, `agent_reminder_schedule`, `agent_media_generate`, and `agent_model_compare` for confirmed delivery, notification, reminder, media generation, and blind model comparison review/judgment.
+- `agent` for visible autonomous work: spawn, batch-spawn, inspect, message, wait, cancel, and report tracked agents. Use `agent_harness mode:"agent_orchestration"` first when deciding whether subagents are useful, reviewing managed milestones, or checking existing agent state.
 
 Registered model tool definitions are compact by default. Top-level descriptions are short, nested parameter descriptions are omitted from the default model catalog, and tool catalog rows include direct harness inspection routes. The model can inspect detailed contracts through `agent_harness mode:"tools"` with `includeParameters:true`, `mode:"tool"`, or the owning harness mode.
 
 ## Local Behavior
 
-Agent-local memory, notes, personas, skills, routines, and profiles are stored under the Agent home. They are injected only into the serial Agent conversation unless an explicit Agent workflow promotes or ingests reviewed material elsewhere.
+VIBE.md is the friendly personality file for GoodVibes Agent. Project and global VIBE.md files are discovered, scanned for secret-looking content, surfaced in setup and the learning curator when blocked or truncated, and applied to the serial Agent conversation without requiring persona-registry ceremony. The direct `vibe` tool exposes status/show plus confirmed init and persona import, while `/vibe` remains the visible command surface. Formal Agent-local memory, notes, personas, skills, routines, and profiles remain stored under the Agent home and are injected only into the serial Agent conversation unless an explicit Agent workflow promotes or ingests reviewed material elsewhere.
+
+Project context files are separate from personality. GoodVibes Agent loads secret-scanned workspace instructions from `.hermes.md`, `HERMES.md`, `AGENTS.md`, `CLAUDE.md`, optional `HERMES_HOME/SOUL.md`, `.cursorrules`, and `.cursor/rules/*.mdc`, with target-aware discovery for subdirectory `AGENTS.md` files. These instructions can shape project work, but explicit user requests, tool contracts, confirmation gates, and safety rules override them.
 
 Useful workspace paths:
 
-- Memory & Skills -> Create memory, Create note, or Capture learned behavior.
+- Memory & Skills -> Memory posture, Create memory, Create note, or Capture learned behavior.
 - Notes -> Create, edit, review, stale, delete, or promote scratchpad notes.
-- Personas -> Create, inspect, activate, review, stale, or delete local personas.
+- Personas -> Inspect, create, show, or import VIBE.md; create, inspect, activate, review, stale, or delete local personas.
 - Skills -> Create skills, import discovered skills, enable/disable, review, delete, and manage skill bundles.
 - Routines -> Create routines, start a routine in chat, review receipts, and explicitly promote one routine to a connected schedule.
-- Profiles -> Create isolated Agent profiles from built-in or imported starter templates.
+- Profiles -> Create isolated Agent profiles from built-in or imported starter templates, with opt-in VIBE.md portability through `--include-vibe`.
 
-Starting a routine prints its steps in the main conversation. It does not start background automation. Promotion to a connected schedule is separate, explicit, confirmation-gated, and records a redacted local receipt.
+Starting a routine prints its steps in the main conversation. Promotion to a connected schedule or automation job is separate, explicit, confirmation-gated, visible in the autonomy queue, and records a redacted local receipt.
 
 ## Knowledge And Artifacts
 
@@ -118,11 +133,11 @@ Agent does not fall back to default knowledge or other product-specific knowledg
 
 The Knowledge workspace and model tools support status, ask/search, source/node/issue lists, item lookup, map review, connector inspection, URL/file/URL-list/bookmark/browser-history/connector ingest, issue review, prompt packet/explain previews, consolidation, review queue, and reindex.
 
-Artifacts are first-class runtime objects for uploaded files, generated media, and delegated outputs. Generated media is stored as GoodVibes artifacts and reported by artifact id and metadata, not inline base64.
+Artifacts are first-class runtime objects for uploaded files, generated media, and delegated outputs. Saved artifacts can be browsed, previewed, attached to drafts, promoted to Agent Knowledge, or exported to workspace files after confirmation. Generated media is stored as GoodVibes artifacts and reported by artifact id and metadata, not inline base64.
 
 ## Connected Host
 
-Start the owning GoodVibes host before launching Agent. The default connection is:
+Connect Agent to a GoodVibes daemon before using daemon-backed features. The default connection is:
 
 ```text
 http://127.0.0.1:3421
@@ -134,17 +149,18 @@ Agent reports unavailable, unauthenticated, or incompatible host state from Agen
 
 Model-visible diagnostics are read-only:
 
-- `service_posture` and `service_endpoint` expose endpoint binding, network-facing posture, issues, optional probes, and redacted log tail.
-- `connected_host` and `daemon` are compact posture maps by default with direct `modelRoute` hints; `includeParameters:true` adds route families, allowed capabilities, blocked lifecycle/non-Agent surfaces, and first-class tool availability. `connected_host_capability` inspects one capability and returns the matching allowed or blocked route hint.
-- `connected_host_status` and `daemon_status` run live readiness checks for host status, host compatibility, token posture, endpoint binding, isolated Agent Knowledge readiness, and the model route to inspect or act on findings.
+- `host action:"services|service"` exposes endpoint binding, network-facing posture, issues, optional probes, and redacted log tail.
+- `host action:"capabilities|capability"` exposes compact posture maps, allowed capabilities, blocked lifecycle/non-Agent surfaces, route families, and first-class tool availability.
+- `host action:"status"` runs live readiness checks for host status, host compatibility, token posture, endpoint binding, isolated Agent Knowledge readiness, and the model route to inspect or act on findings.
+- `setup action:"repair"` selects the next safe setup repair route without executing it: token repair, connected-host status, services.status receipt, user-run bootstrap commands, or no lifecycle action when the host is reachable.
 
-Host lifecycle stays outside GoodVibes Agent.
+Service lifecycle and listener changes are available only through explicit setup or confirmed daemon operator routes when the connected daemon supports them. Raw danger toggles stay protected.
 
 ## Product Boundary
 
-GoodVibes Agent owns the operator assistant TUI, serial assistant flow, local behavior libraries, Agent Knowledge routes, companion chat, work-plan tracking, approvals/automation observability, channel/media/reminder tools, and explicit build delegation.
+GoodVibes Agent owns the autonomous assistant harness: the terminal renderer, setup UX, chat, local behavior libraries, Agent Knowledge routes, companion chat, visible agents, work plans, schedules, approvals, channel/media/reminder tools, daemon method access, and user-facing autonomy status.
 
-GoodVibes TUI owns coding execution: file edits, git/worktree workflows, coding panels, execution isolation UX, and delegated review execution. Agent may delegate explicit build/fix/review work; normal assistant chat does not use shared coding sessions.
+GoodVibes TUI remains the vibecoding harness and is still useful when the user wants its coding panels or execution-isolation UX. Agent should use local tools, visible agents, daemon automation, delegation, or remote runners according to what best serves the user, not because package boundaries are exposed as product friction.
 
 ## Package Docs