v1.1.7

What's Changed

• feat(widgets): entity card for orgx_inspect, Work Ledger for the chronicle; retire daily-brief by @hopeatina in #402
• feat(tools): orgx_lease so agents claim files before editing by @hopeatina in #403
• feat(auth): verify run tokens v2 and forward run delegation claims by @hopeatina in #404
• fix(decide): default a decision's workspace for sessions that never bootstrapped by @hopeatina in #405
• feat(spawn): one runtime choice for OrgX-controlled work by @hopeatina in #406
• orgx_search scope=work_ledger: search Agent Work Receipts as a work graph by @hopeatina in #407
• One dispatch contract behind all four dispatch verbs by @hopeatina in #408
• Bring error, success, and the homepage onto consent's design by @hopeatina in #409
• One acceptance verdict, in the app by @hopeatina in #410
• fix(auth): a v2 run token gets only the tools it was granted by @hopeatina in #412
• fix(api): never re-send an unkeyed write to the fallback after the primary by @hopeatina in #413
• Live state across the widgets: one subscription layer, one work graph by @hopeatina in #411
• fix(dispatch): a spawn without a task is still a durable delegation by @hopeatina in #414
• fix(spawn): conform orgx_spawn output to its published schema by @hopeatina in #416
• fix(spawn): repair a spawn payload that does not match its schema instead of failing by @hopeatina in #417
• Name the real effect vocabulary in the dispatch schema by @hopeatina in #415
• test(live): one harness, one home per subject, behaviour over source greps by @hopeatina in #418
• Stop the widget manifest churning, drop the stale homepage mirror by @hopeatina in #419
• feat(inspect): orgx_inspect(type: run) returns what the run produced by @hopeatina in #420
• Harden MCP admission, analytics privacy, and plan response compatibility by @hopeatina in #424
• Deliver real errors to MCP clients; output schemas for all 41 v2 tools by @hopeatina in #421
• fix(widgets): restore base styling three widgets never inherited by @hopeatina in #423
• fix(workspace-map): stop showing the agent's routing table to the person by @hopeatina in #425
• Fix live MCP output contracts and stop widgets masking failures by @hopeatina in #426
• Accept degraded briefs and bound status previews with inspection links by @hopeatina in #427
• Preserve recommendation data and disclose partial widget reports by @hopeatina in #428
• Never report a widget action as done when the host did not do it by @hopeatina in #429
• Approve ordinary decisions from the decisions widget; send the rest to OrgX by @hopeatina in #430
• Add orgx_command_status so widgets can see what happened by @hopeatina in #431
• Serve the agent avatar renders for widgets and plugins by @hopeatina in #432
• Bring the OrgX design kit into the widgets; decisions shows real avatars by @hopeatina in #433
• Search, entity card, workspace map and Work Ledger on the design kit by @hopeatina in #434
• Add the OrgX panel for the ChatGPT sidebar and beside a thread by @hopeatina in #435
• Task spawned, scaffolded initiative, plan session and artifact review on the design kit by @hopeatina in #436
• Fix search response delivery and add widget failure diagnostics by @hopeatina in #437
• Fix search card recovery, filters, readable titles, and help by @hopeatina in #438
• Decisions, morning brief, agent status and initiative pulse on the design kit by @hopeatina in #440
• Let agent profiles check what happened to started work by @hopeatina in #441
• Filter full MCP searches by creation date before result limits by @hopeatina in #439
• Single-click approvals, lime primaries, and decision buttons from the server by @hopeatina in #442
• Show the last accepted output from the app's acceptance ledger by @hopeatina in #443
• Receipt explorer, proof ladder and model mark from the design canvas by @hopeatina in #444
• Inline only the kit bundles each widget uses, and budget every widget's payload by @hopeatina in #445
• Settle every decision type in place from the decisions widget and the panel by @hopeatina in #447
• Run three legacy tool names as thin aliases of their v2-core tools by @hopeatina in #446
• Bring the weekly widget contract checks back to green, and show widgets under reduced motion by @hopeatina in #448
• Fix approval-list output validation and panel touch target by @hopeatina in #449
• Fix live ChatGPT QA failures in the pulse, search, agent status and scaffold widgets by @hopeatina in #450
• Fix live ChatGPT QA failures in the decision widgets: readable errors, short headlines, honest states, one focus ring by @hopeatina in #451
• One public tool contract per profile, callable widget tools, needs_human decisions and a CI gate against drift by @hopeatina in #452
• Accept the app's pending-decision scope, brief source health and review action before it ships by @hopeatina in #453
• Widget UX foundations: human headshots with hover cards, local time, working OrgX links, icons, surfaces, smaller payloads by @hopeatina in #454
• One type scale without mono eyebrows; entity card, work ledger, workspace map, proof receipt, streaming scaffold and gallery reworked by @hopeatina in #456
• Decision widgets: whole headlines, structured bodies, a plain quality verdict, honest scope and no jargon by @hopeatina in #457
• Pulse, search, agent status, scaffold and plan: real progress, usable filters, hover cards in rows, no id chips by @hopeatina in #458
• Widget cleanup: footers that wrap, lighter panel queue, animated clamps, honest fallback counts, server links on real routes by @hopeatina in #459
• Clarify approval review titles and audit OpenAI held MCP updates by @hopeatina in #461
• Prepare Claude directory tool metadata for submission by @hopeatina in #462
• Expand Claude directory to the full OrgX workflow surface by @hopeatina in #463
• Require run ownership before MCP Resume by @hopeatina in #464
• MCP broker gateway: /c/ pass-through with run tokens, tool filtering and metadata-only logs by @hopeatina in #466
• Verify run tokens with the dedicated secret or the service key; sync run-token and broker secrets after deploy by @hopeatina in #467
• Forward artifact idempotency keys without enabling write retries by @hopeatina in #469
• Keep panel decisions scoped through selection and settlement by @hopeatina in #468
• Let the run-token service-key fallback expire on its own by @hopeatina in #472
• Harden the MCP broker proxy: forward only the tool surface, own-key lookups, fail-closed filtering by @hopeatina in #473
• OrgX panel v2: whole queue, tabs, receipts, start-work prompts and a first-use tour by @hopeatina in #471
• OrgX panel: live updates from a per-viewer feed by @hopeatina in #474
• test(panel): fix announcement timing race in live integration test by @hopeatina in #476
• OrgX panel: switch workspaces from the panel by @hopeatina in #475
• OrgX panel: polish pass from the production panel in ChatGPT by @hopeatina in #477
• feat(panel): Start tab, drill-down In progress and Done, honest links by @hopeatina in #478
• feat(panel): show the work and its verdicts; a person's call on receipts by @hopeatina in #480
• orgx_submit_receipt: record the model, pass full receipts through verbatim by @hopeatina in #479
• fix(panel): survive ChatGPT's saved copy of the tool schema by @hopeatina in #482
• Agree on what done means before work starts by @hopeatina in #484
• Name the model once at orgx_bootstrap; receipts default to it by @hopeatina in #483
• Move the panel's preview gallery out of the inlined payload by @hopeatina in #481
• security: patch dependencies and pin CI actions by @hopeatina in #485

Full Changelog: v1.1.6...v1.1.7