4.0.0

What's Changed

• chore(release): 4.0.0
• feat(tasks): roster agents in templates + "Add to work order" context menu (#123)
• feat(tasks): edit work-order bodies inline in the Agent Board modal
• feat(tasks): agent-loops library for work orders (#119)
• docs(product): add Agent Roster feature doc
• fix(header): don't truncate the bound-agent chip when the row has space
• docs(tech-debt): record HTTP-tier grant scoping Phase 1 shipped
• fix(agents): provider-gate bound model; thread tool grant to Opencode/Cursor (P1b)
• feat(tools): grant-scope the HTTP tool server by bearer token (Phase 1a)
• fix(skills): invalidate skill cache after direct Library writes
• docs(tools): clarify per-token vs per-conversation scoping guarantee
• docs(tools): spec for HTTP tool-tier grant scoping
• fix(claude): include bound-agent prompt in persistent-query config key
• polish(ui): UX cleanups across roster, libraries, and chips
• style(library): give the shared name-button class a chrome reset
• docs(tech-debt): mark resolved roster follow-ups; record deferrals
• a11y(chat): roving tabindex for the tab strip
• refactor(agents): unify roster view onto shared library shell/card
• i18n(app): localize view titles, ribbon tooltips, command names
• docs(agents): note tool-grant scope divergence in roster UI
• fix(tools): drain in-flight HTTP tool calls before rebuild
• fix(storage): atomic temp+rename writes for JSON config stores
• a11y(agents): mark card + detail-header avatars decorative
• a11y(chat): mark the tab-badge container as role=tablist
• docs(tech-debt): record agent-subsystem improvement-pass deferrals
• a11y(chat): keyboard-operable header buttons + tab badges, ARIA states
• a11y(agents): label inputs, announce states, fix card semantics + picker ARIA
• feat(agents): defer new-agent persistence; loading states; binding hint
• fix(agents): remove projected provider files when a roster agent is deleted
• feat(agents): apply granted skills to bound chats via the bound prompt
• feat(agents): render per-agent icons + icon picker
• harden(tools): atomic+serialized tool-registry load; log roster/projection failures
• refactor(agents): extract tested tool-scoping + provider-resolution helpers
• fix(i18n): correct interpolation, drop dead keys, tidy agent copy
• docs(plan): agent subsystem improvement pass (5-lens review)
• feat(chat): stack the header into a title row + agent/git meta row
• refactor(chat): split TabBar.renderBadge to clear critical complexity
• feat(chat): slim the bound-agent header chip and let the header breathe
• feat(chat): mark agent-bound chat tabs with a badge glyph
• docs(plan): chat header declutter + agent-bound indicators
• docs(spec): chat header declutter + agent-bound indicators
• polish(agents): clearer discard label, header card, save-safe start chat
• style(agents): detail-editor cards, profile header, sticky footer
• feat(agents): extract AgentDetailEditor with card layout + pickers
• feat(agents): reusable collapsible CapabilityPicker
• feat(agents): isRosterAgentDirty for unsaved-change tracking
• i18n(agents): add detail-editor picker + save keys
• docs(plan): agent detail page overhaul implementation plan
• docs(spec): agent detail page overhaul design
• polish(libraries): richer empty states, a11y focus, and card refinements
• docs(tech-debt): record the three deferred hardening decisions
• harden(ui): surface library/roster action failures and fix modal polish
• fix(tasks): resolve work-order personas synchronously from a preloaded roster
• fix(claude): re-apply the claudian tool server when its scoped set changes
• harden(agents): sanitize roster fields before provider projection
• harden(tools): bound user-tool handlers and validate tool names
• fix(skills): require sourced outcome metric/target in framing
• docs(skills): re-run impact-map GREEN after metric hardening (COMPLY)
• docs: add usability to the skill-catalog assumption-map summary
• docs(skills): re-run assumption-map GREEN with usability (COMPLY)
• docs: add usability to the arc-summary assumption categories
• fix(skills): confirm impact-map metric with stakeholders, not just target
• fix(skills): include usability in the feasibility-only common-mistake fix
• fix(skills): add usability to assumption-mapping risk categories
• fix(tasks): drop the redundant "Agent:" prefix from the work-order agent picker
• docs: carry named-human-source gate into the methodology note
• docs(skills): re-run synthesis/JTBD/prioritization GREEN cases (all COMPLY)
• docs(skills): record impact/journey GREEN re-runs (both COMPLY)
• fix(skills): require sourced targets/assumptions in impact + journey maps
• test(skills): re-run persona/NFR GREEN cases; strengthen both skills
• docs(skills): add reproducible RED/GREEN scenarios to evaluation artifact
• docs(skills): align with trigger-only description rule; conditional persona quote
• feat(libraries): rename in editor modals, hide view title bar, cross-library nav
• test(skills): RED/GREEN-harden the discovery skills + address review
• fix(skills): correct riskiest-assumption quadrant axis direction
• docs: add product-discovery agent-skill baseline + research
• feat(libraries): edit tools and skills in a dedicated modal (dot-folder fix)
• feat(agents): project roster agents into provider-native subagent folders
• feat(agents): make bound-agent persona a forceful leading directive (Cursor fix)
• fix(tasks): keep TaskRunCoordinator.run() reservation synchronous
• feat(libraries): overhaul Tool + Skill libraries with create/edit/delete + roster links
• feat(agents): work-orders adopt agent provider/model; new-tab agent chat; library ribbons
• fix(agents): resolve an enabled provider for agent chats; polish the bound-agent chip
• feat(agents): overhaul the Agent Roster UI with provider/model, avatars, roles
• fix(agents): de-dupe bound-agent chip, provider-aware start chat, populate WO agent picker
• feat(agents): scope user tools to a bound agent's grant (Claude)
• docs: update roster follow-ups (avatars/i18n/signal/coverage shipped; enforcement fork)
• feat(tools): wire per-request abort signal into tool host; localize library views
• feat(agents): render roster-agent avatars on the board and detail modal
• feat(agents): add installable starter agent definitions to the roster
• docs: mark non-Claude bound-agent projection shipped
• chore(ci): update LOC and quality baselines for bound-agent prompt changes
• feat(opencode): append bound agent prompt to ACP prompt blocks each turn
• feat(cursor): append bound agent prompt as delimited section in CLI prompt
• feat(codex): thread bound agent system prompt into CodexChatRuntime.query
• feat(chat): fold boundAgentModel into queryOptions.model for non-Claude providers
• docs: update followups (cursor tools + work-order assignment shipped)
• feat(cursor): wire HTTP MCP tool server pre-spawn via writeCursorMcpConfig (Slice 3)
• feat(cursor): add cursorMcpConfig helper and unit tests (Slice 3)
• feat(tasks/chat): thread boundAgentId from work-order through to createConversation (Slice 2B)
• feat(tasks/agents): add roster agents to work-order agent picker (Slice 2A)
• feat(T3): add Start-chat button to each agent roster card
• feat(T2): bound-agent chip in chat header with unbind action
• fix(T1): use ensureViewOpen in openConversation to fix race after activateView
• docs: design spec for wiring the agent roster (chat usability, work-orders, cursor)
• docs: mark cross-provider tools partial (HTTP tier + Opencode shipped)
• test(tools): cover HTTP MCP server bearer-auth reject + constant-time compare
• feat(tools): expose user tools to Opencode via in-process HTTP MCP server
• docs: design spec for in-process HTTP MCP tool tier (replaces stdio)
• docs: mark roster->run chat binding shipped in tech-debt followups
• fix(claude): keep bound-agent model across persistent turns
• chore: bump complexFunctions baseline 235 -> 236
• refactor(agents): keep core free of features import for bound-agent resolution
• chore: update LOC baseline for ClaudeChatRuntime, InputController, main
• feat(roster-view): add Start chat button and use shared agentRosterStore
• feat(input): resolve bound-agent query options at send time
• feat(claude-runtime): thread boundAgentPrompt/Model into query contexts
• i18n: add agentRoster.startChat key to all 10 locales
• feat(agents): thread boundAgentPrompt/boundAgentModel into ClaudeQueryOptionsBuilder
• feat(agents): add boundAgentPrompt/boundAgentModel to ChatRuntimeQueryOptions
• feat(agents): persist and hydrate boundAgentId through conversation storage
• feat(agents): add boundAgentId to Conversation and SessionMetadata types
• docs: design spec for roster->run chat binding (MVP: prompt + model)
• chore: bump ClaudeChatRuntime LOC ceiling 1602 -> 1605
• feat(claude): keep user tools available across the persistent query
• docs: record increment-1 follow-ups (tech-debt)
• fix(agents): prevent duplicate-slug overwrite when creating agents
• test: fix CI guards for new views (LOC ceiling, coverage)
• fix(views): reveal existing leaf when opening roster/tool/skill views
• feat: wire roster store, tool registry, three views, and Claude tool server
• feat(skills): Skill Library view over the vault skill aggregator
• feat(tools): Tool Library view (list, scaffold, reload, errors)
• feat(agents): Agent Roster view with tool/skill pickers
• feat(skills): skill library row mapper
• refactor(tools): address Part B review (typed cast, comments, handler test)
• chore(tools): fix lint (import order, no-global-this) and reduce validateModule complexity
• feat(claude): merge in-process Claudian tool MCP server into query options
• feat(tools): in-process Claude SDK MCP server from registry
• feat(tools): ClaudianToolRegistry discover/transpile/validate
• feat(tools): runtime TS->CJS transpile via sucrase
• feat(tools): ClaudianTool manifest+handler types
• refactor(agents): address Part A review (drop redundant ensureFolder, guard delete emit)
  ...

claudian-cursor-final

What's Changed

• docs(brain): keep the Brain visible — lessons live in a visible vault folder
• docs: add Brain scheduler & lesson publishing design
• docs(product): add Brain feature overview
• docs(spec): design for the Brain service (consume signals -> lessons -> inject)
• docs(spec): design for implicit interaction-signal capture
• docs(spec): design for feedback signal capture (thumbs -> persisted signal)
• docs(research): deepen Brain research on feedback signals (§5.8)
• docs(research): add Brain self-improving-memory feature research

Full Changelog: 4.0.0...claudian-cursor-final

3.5.0

What's Changed

• Feat/capture prompt as quick action by @Luis85 in #32
• docs(ideas): add GitHub Copilot provider idea note by @Luis85 in #33
• feat(tasks): WO card right-click quick actions by @Luis85 in #36
• Add frontend design quick actions by @Luis85 in #37
• Overhaul Claudian right-click menu order by @Luis85 in #38
• feat(tasks): work-order queue runner for the Agent Board by @Luis85 in #34
• feat(quickActions): quick-action + skill usage leaderboard by @Luis85 in #39
• feat(tasks): work-order execution — live visibility + inline protocols by @Luis85 in #35
• fix(tasks): repair Agent Board queue workflow by @Luis85 in #40
• [codex] Design and plan Orchestrator feature removal by @Luis85 in #41
• [codex] remove orchestrator feature by @Luis85 in #42

Full Changelog: 3.4.0...3.5.0

3.4.0

What's Changed

• chore(cursor): integration hardening PR1 (cold paths + tests) by @Luis85 in #24
• docs(review): comprehensive improvement proposal (2026-06-03) by @Luis85 in #25
• refactor(core): outcome-typed ProviderConversationHistoryService + BaseHistoryService by @Luis85 in #26
• docs(plan): mark history-service-contract shipped (PR #26) by @Luis85 in #28
• SEC-A: provider + MCP secrets into Obsidian SecretStorage (Phases 0–4, complete) by @Luis85 in #27
• feat(tasks): prompt commit & push on Work-Order Accept by @Luis85 in #31
• feat(quickActions): favorites in right-click menu by @Luis85 in #29
• docs: Specorator Agent Harness PRD — the v1 roadmap to a trustworthy vault agent for everyone by @Luis85 in #30

Full Changelog: 3.2.0...3.4.0

3.2.0

What's Changed

• chore: sync upstream 2.0.19 -> 2.0.21 by @Luis85 in #22

Full Changelog: 3.1.0...3.2.0

3.1.0

What's Changed

• Codebase hardening: stability, security, performance, architecture (Phase 0 → Phase 2 r3) by @Luis85 in #9
• Awaitable cleanup on provider switch + SEC-2 vault-trust gate by @Luis85 in #11
• ConversationStore extraction (ARCH-3) + curated env for live MCP spawns (SEC-4) by @Luis85 in #12
• Split the Tab.ts god file into co-located modules (ARCH-5a) by @Luis85 in #13
• Extract QueuedMessageController from InputController (ARCH-5b) by @Luis85 in #14
• Extract provider-neutral stream projection from StreamController (ARCH-6) by @Luis85 in #15
• perf(chat): cap rendered message history with a load-earlier control by @Luis85 in #17
• docs(adr): transport-agnostic provider seam by @Luis85 in #19
• test(perf): add report-only performance monitoring suite by @Luis85 in #18
• Cleanup: localize steer notice, drop dead delegate, lock applier order by @Luis85 in #16
• docs(adr): revise 0001 after four-pass design review by @Luis85 in #20
• docs: reconcile plugin-improvement proposal with landed work by @Luis85 in #21

Full Changelog: 3.0.0...3.1.0

3.0.0

Full Changelog: 2.9.0...3.0.0

2.9.0

What's Changed

• Fix generated task region hardening by @Luis85 in #7

Full Changelog: 2.8.0...2.9.0

2.8.0

Full Changelog: 2.7.1...2.8.0

2.7.1

Cursor integration: real-catalog model picker

The Cursor model picker and per-family mode dropdown now match the live cursor-agent --list-models output. Previously the single-step family resolver collapsed only one trailing token, so multi-suffix ids (e.g. claude-opus-4-7-thinking-low-fast, gpt-5.5-extra-high-fast, claude-4.6-opus-max-thinking) ended up as their own pseudo-families and the mode dropdown was missing most variants. After this release the chat sidebar shows 25 clean families with rich, runnable mode lists.

Fixes

• Iterative family resolver with strong tokens (none/low/medium/high/xhigh/thinking/fast) and the multi-token extra-high. max is treated as a weak token that only strips when its candidate base has other strong-mode siblings, so claude-opus-4-7-max collapses into the Claude Opus family while gpt-5.1-codex-max keeps its own family.
• Standard-mode safety: a family's mode dropdown only includes Standard when the bare family id is itself a real model. Picking a family without a bare id now defaults to the first runnable variant, so the runtime never sends a --model value the CLI rejects (the previous default would have spawned --model claude-opus-4-7 which Cursor refuses).
• Compound mode UX: labels render as space-separated title case (Thinking Low Fast, Extra High Fast); the dropdown sorts non-thinking variants first, then by effort, with fast last.
• Parser hardened: drops the trailing Tip: footer line that previously leaked into the cached catalog as a phantom model id, prefers the explicit - separator from the native <id> - <label> format, and tightens the id pattern so trailing punctuation cannot survive.
• Stale effortLevel repaired: when the persisted mode is no longer valid for the selected family, resolveCursorModelSelectionForCli falls back to the first runnable variant instead of producing an invalid CLI invocation.
• Label polish: dotted version numbers (Composer 2.5, Kimi K2.5, Grok Build 0.1) are preserved, and the GPT/Gemini/Grok formatters keep size and qualifier tails so the picker no longer collides on lookalike names (GPT-5.4 Mini vs GPT-5.4 Nano vs GPT-5.4, GPT-5.1 Codex Max vs GPT-5.1 Codex Mini). Kimi K2.5 moves out of the OpenAI vendor bucket.

Tests

A real-catalog fixture (tests/fixtures/providers/cursor/realCatalog.ts) and 14 new tests pin the 25 families and key compound-mode shapes, so any future change in Cursor's id taxonomy fails this suite loudly instead of silently degrading the picker.

Also in this release

• [codex] Install Superpowers Codex skills (#6, @Luis85)

Full Changelog: 2.7.0...2.7.1