Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Generated SDK contracts + publishing groundwork (Phase A) #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Manifest v2, permissions, config consolidation + hook source (Pha…

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Focused Phase B gates passed; full make verify remains reserved for final program completion.

Co-Authored-By: Codex <[email protected]>

* feat: Code-first toolchain: describe mode, build/validate/init (Phase C…

Checkpoint via cy-loop-tasks (iteration 4, phase B mode=tasks).

Scoped Phase C gates passed; the single monorepo make verify remains reserved for final program completion.

Co-Authored-By: Codex <[email protected]>

* feat: Dev lane: links, instance-keyed reload, logs, watch (Phase D) #4

Checkpoint via cy-loop-tasks (iteration 5, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Distribution: source union, gitsrc, sidecars, publish, search, up…

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Contributed commands surface (ADR-008, fixture-proven) #6

Checkpoint via cy-loop-tasks (iteration 7, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: CLI UX & operability: diagnostics, status, doctor, events matrix…

Checkpoint via cy-loop-tasks (iteration 8, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Web surfaces: update affordance, install forms, dev badge, logs p…

Checkpoint via cy-loop-tasks (iteration 9, phase B mode=tasks).

Visual recapture was explicitly waived by the operator; functional browser and focused gates passed.

Co-Authored-By: Codex <[email protected]>

* feat: Docs, official skill, examples, QA scenario flags (Phase G) #9

Checkpoint via cy-loop-tasks (iteration 10, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: QA Plan and Session Charters #10

Checkpoint via cy-loop-tasks (iteration 11, phase B mode=tasks).

QA planning checkpoint: six targeted extension charters, one Marketplace/Skills canary, 676 materialized scenarios, and explicit coverage of Safety Invariants 1-17 plus ADRs 001-008.

Co-Authored-By: Codex <[email protected]>

* fix: close extension QA defects

* fix: remediate extension deep review

* fix: review round

* fix: resolve post-rebase integration defects

---------

Co-authored-by: Codex <[email protected]>

* feat: Schema, classification, DSL grammar and config contracts #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: implement loop node failure precedence

* feat: Effects: outbox, relay and delivery identity #3

Checkpoint via cy-loop-tasks (iteration 4, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Target health breaker, quarantine and requeue #4

Checkpoint via cy-loop-tasks (iteration 5, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* fix: resolve CodeRabbit lifecycle review

* fix: address incremental CodeRabbit feedback

---------

Co-authored-by: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Tauri shell: windows, native integration, runtime resolution #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: add desktop app CLI and runtime contracts

* feat: Update system: provisioning, runtime apply, app auto-update, cont…

Checkpoint via cy-loop-tasks (iteration 4, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Product language + docs: brand hard cut, site pages, official ski…

Checkpoint via cy-loop-tasks (iteration 5, phase B mode=tasks).

Hard-cut public product language to CompozyOS, ship desktop documentation and official-skill routing, and enforce the boundary with UT-113.

Co-Authored-By: Codex <[email protected]>

* feat: Release pipeline: build matrix, signing, feeds, gates, custody ru…

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: qa-report: plan the desktop QA cycle in docs/qa #6

Checkpoint via cy-loop-tasks (iteration 7, phase B mode=tasks).

Planning-only QA checkpoint: 13 APP scenarios, five desktop journeys, and 17 isolated per-platform charters. No QA lab or product process was started.

Co-Authored-By: Codex <[email protected]>

* fix: initialize the desktop updater in development

* fix: start the desktop runtime through the current daemon contract

* fix: support runtime launchers with child processes

* fix: bind desktop probes to runtime identity fields

* fix: accept the current beta runtime in desktop handshakes

* fix: frame desktop control requests by newline

* fix: make desktop recovery state safe to retry

* test: verify desktop release candidate

* fix: peer-review round 1 remediation

Checkpoint via cy-loop-tasks (iteration 10, phase D review round 1).

Co-Authored-By: Codex <[email protected]>

* fix: preserve product language after main rebase

* fix: review round

* build: desktop dev command

---------

Co-authored-by: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Focused evidence: worktree race unit suite, real-Git lifecycle integration, migration integrity/equivalence, and 80.8% package coverage. Heavy gates remain deferred to the QA tail by workflow direction.

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Focused evidence: worktree race unit suite, real-Git lifecycle integration, migration integrity/equivalence, and 80.8% package coverage. Heavy gates remain deferred to the QA tail by workflow direction.

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Focused evidence: worktree race unit suite, real-Git lifecycle integration, migration integrity/equivalence, and 80.8% package coverage. Heavy gates remain deferred to the QA tail by workflow direction.

Co-Authored-By: Codex <[email protected]>

* feat: Schema and worktree domain core #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Focused evidence: worktree race unit suite, real-Git lifecycle integration, migration integrity/equivalence, and 80.8% package coverage. Heavy gates remain deferred to the QA tail by workflow direction.

Co-Authored-By: Codex <[email protected]>

* feat: Worktree public surface: API, CLI, native tools, streams #2

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Session binding, containment, spawn inheritance, fork #3

Checkpoint via cy-loop-tasks (iteration 4, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Environment policies: task worktree policy + loop environment #4

Checkpoint via cy-loop-tasks (iteration 5, phase B mode=tasks).

Validated task and Loop worktree placement across queue snapshots, lease-fenced session binding, public contracts, real Git journeys, and generated artifacts.

Co-Authored-By: Codex <[email protected]>

* feat: Assisted exit, forge.provider surface, bundled GitHub extension #…

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: add web worktree navigation and lifecycle

* feat: Web: environment surfaces and assisted exit #7

Checkpoint via cy-loop-tasks (iteration 10, phase B mode=tasks).

Focused Go race suites, web tests, typecheck, and lint passed. Browser and visual evidence remain assigned to the QA tail.

Co-Authored-By: Codex <[email protected]>

* docs: document Worktree operations and agent surfaces #8

* feat: QA Plan and Session Charters #9

Checkpoint via cy-loop-tasks (iteration 14, phase B mode=tasks).

Adds the durable Worktree QA plan and closes lint, optional-state, native-tool metadata, and bundled-skill inventory gaps exposed by the full gate.

Co-Authored-By: Codex <[email protected]>

* fix: preserve operator global config at daemon boot

* fix: mount desktop shell context before projection reads

* docs: record autonomous observer mismatch replay

* fix: preserve pending worktree creation state

* fix: expose worktree config lifecycle

* fix: reject invalid worktree bases before acceptance

* fix: preserve task fanout intent identity

* fix: start native task claims in bound sessions

* test: complete worktree real-user QA

* fix: keep worktree exit menu mounted

* fix: complete worktree review remediation

* fix: reconcile main rebase contracts

* fix: lint

* docs: update design

* fix: a lot of adjustments

---------

Co-authored-by: Codex <[email protected]>

* feat: Conformance foundations: mcppolicy + agentplugin #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Focused race tests, scoped lint, test conventions, package boundaries, and coverage passed. Broad gates remain deferred to the QA/review tail.

Co-Authored-By: Codex <[email protected]>

* feat: ingest portable agent plugins

* feat: add portable MCP runtime bindings

* feat: expose portable extension diagnostics

* feat: Web + catalog: format badges, skipped components, feed hard cut #…

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* docs: document agent plugin interoperability

* docs: plan agent plugin QA

* fix: resolve agent plugin review findings

* docs: close agent plugin QA findings

* fix: use canonical product name

* refactor: split CLI validation error rendering

* docs: format agent plugin reference

* fix: align agent plugin verification contracts

* docs: narrow agent plugin provider claim

* fix: resolve agent plugin review findings

* fix: allow secure agent plugin file reads

* fix: preserve agent plugin review invariants

* fix: satisfy agent plugin lint invariants

* fix: restore portable MCP e2e contract

* fix: resolve portable plugin fixture path

* fix: align portable validation transcript

---------

Co-authored-by: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Update Operation, single command, and daemon surfaces #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Web update surface (settings two-track + menubar indicator) #2

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Claude Opus implementation reviewed and accepted by the controller; visual and browser QA evidence remains owned by the tail tasks.

Co-Authored-By: Codex <[email protected]>

* feat: Electron shell at parity #3

Checkpoint via cy-loop-tasks (iteration 4, task mode).

Co-Authored-By: Codex <[email protected]>

* feat: add desktop release authority

* feat: Cutover and docs truth pass #5

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Hard cut the legacy desktop shell and feed, retire the external bucket and alias, and align docs, skill, QA tracking, and release guidance with Electron.

Co-Authored-By: Codex <[email protected]>

* feat: QA Plan and Session Charters #6

Checkpoint via cy-loop-tasks (iteration 7, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* fix: remediate electron shell review findings

* fix: scope desktop notarization to macOS

* fix: use configured macOS notarization credentials

* fix: run desktop release builder from app workspace

* fix: execute release builder in desktop workspace

* fix: harden desktop release artifact contracts

* fix: complete Linux and DMG package metadata

* fix: preserve desktop release identity

* fix: normalize signed x64 runtime digest

* fix: synchronize web assets for explicit releases

* fix: pass release fallback to web assets sync

* build: sync web assets for explicit release

* fix: refresh finalized dmg integrity

* build: sync web assets for explicit release

* build: sync web assets for explicit release

* fix: exclude dmg files from update channel

* build: sync web assets for explicit release

* fix: exclude deb files from update channel

* build: sync web assets for explicit release

* fix: isolate desktop release smoke

* docs: record electron shell QA closeout

* fix: final adjusts

* fix: update icon

* fix: close buttons

* fix: resolve pull request lint failures

---------

Co-authored-by: Codex <[email protected]>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Focused verification: 4,604 tests passed; full gate deferred to the QA tail.

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Focused verification: 4,604 tests passed; full gate deferred to the QA tail.

Co-Authored-By: Codex <[email protected]>

* feat: Source config foundation and policy write path #1

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Focused verification: 4,604 tests passed; full gate deferred to the QA tail.

Co-Authored-By: Codex <[email protected]>

* feat: add live skill source discovery

* feat: add transactional skill exposures

* feat: expose skills across public surfaces

* feat: add provider-aware skill sessions

* feat: add skill source web surfaces

* docs: document skill sources and exposure

* fix: support profile skill source filters

* fix: resolve skill sources review findings

* docs: plan skill sources QA sessions

* fix: accept skill source overrides at exact-profile scope

BUG-20260825-skill-source-profile-write-rejected

Setting or unsetting skills.sources / skills.custom_sources for a named
profile failed with `decode skills settings request: unknown field
"override"` on every transport, leaving the profile layer of the
four-layer source contract unreachable.

internal/settings.updateSkillsSection already accepts a
SkillSourcesOverride at ScopeProfile, and the CLI already sends it —
the presence-aware override is what gives `config unset` its
clear-and-inherit semantics. Only the shared API decoder was gated on
ScopeWorkspace, so profile bodies fell through to the config-only
branch and were rejected.

Offer the override shape to the exact-profile lane too, keeping the
workspace lane override-only and the user lane config-only. The
forbidden-field refusal keeps its `workspace_scope_field_forbidden`
code and now words its message for whichever scope raised it.

Co-Authored-By: Claude Opus 5 <[email protected]>

* docs: correct the agent write policy for skill source keys

BUG-20260825-skill-source-agent-write-doc-mismatch

The official skill told agents that skills.sources and
skills.custom_sources are trust roots which compozy__config_set and
compozy__config_unset deny with config_trust_root_forbidden. Both tools
write both keys at user and workspace scope and report applied: true,
lifecycle: live; agent and profile scope refuse with
config_scope_not_allowed, and config_trust_root_forbidden is never
emitted for either key.

The runtime matches the binding spec, so the references were wrong.
Both paths sit in agentMutableConfigKinds and ClassifyToolConfigPath
returns on that lookup before the trust-root branch; the keys also
appear in skillsConfigPathIsTrustRoot, which is the unreachable branch
the original sentence was written from.

Correct both sentences and pin the classifier outcome in the canonical
tool-surface policy table so a future reordering cannot silently make
the corrected text wrong again. The missing doc-vs-runtime gate is
recorded in the QA automation backlog.

Co-Authored-By: Claude Opus 5 <[email protected]>

* fix: accept the registered workspace id on skill detail and expose routes

BUG-20260825-skill-detail-rejects-workspace-id

compozy skill info, skill where, skill expose and skill unexpose
refused every --workspace value, including the exact canonical ws_ id
workspace info prints, with "workspace_id must be the canonical
workspace id". GET /api/skills/{name} and the expose routes failed the
same way on HTTP and UDS, so no client could work around it.

canonicalResolvedWorkspaceID preferred ResolvedWorkspace.WorkspaceID,
which the resolver stamps with the durable identity from
<root>/.compozy/workspace.toml, over ResolvedWorkspace.ID, which holds
the registered ws_ id every public surface emits. The comparison
measured the caller's public id against an identity no public surface
ever hands out. The failing expose envelope echoed the public id at the
top level while rejecting that same string.

Prefer the registered id and fall back to the durable identity only for
a workspace resolved by path that has no registered id.

The existing coverage passed because its stub ResolvedWorkspace left
WorkspaceID empty, a shape production never has; the new subtest builds
both fields the way the resolver does.

Co-Authored-By: Claude Opus 5 <[email protected]>

* fix: record every effective skill source root in apply events

* fix: reject non-source skill fields at workspace scope

* fix: report untouched exposure targets accurately

* fix: close skill source verification gaps

* fix: preserve global profile resource loading

* fix: project skills by registered workspace id

* fix: close skill source QA regressions

* fix: complete skill source QA closure

* build: update deps

* fix: preserve contracts after main rebase

---------

Co-authored-by: Codex <[email protected]>
Co-authored-by: Claude Opus 5 <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

#490)

* feat: Terminal core and PTY substrate #1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Wire protocol and transport #2

Checkpoint via cy-loop-tasks (iteration 3, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Persistence, journal and observability #3

Checkpoint via cy-loop-tasks (iteration 4, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Agent surfaces exec tools and write path #4

* feat: Web terminal system #5

Checkpoint via cy-loop-tasks (iteration 6, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: Public activation tranche #6

Checkpoint via cy-loop-tasks (iteration 7, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* feat: render agent-reported terminal output

* feat: Windows ConPTY adapter #8

Checkpoint via cy-loop-tasks (iteration 9, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* fix: satisfy integrated terminal gate policies

* fix: normalize local gate classification

* fix: resolve integrated terminal review findings

* feat: QA Plan and Session Charters #9

Checkpoint via cy-loop-tasks (iteration 10, phase B mode=tasks).

Co-Authored-By: Codex <[email protected]>

* fix: preserve terminal workspace registration identity

* fix: complete integrated terminal QA remediation

* refactor: remove unused terminal exports

* fix: ui improvements

* fix: improvs

* fix: complete rebase integration

* fix: align product language

* fix: resolve Windows terminal executables

* test: stabilize Windows terminal reads

* test: drain Windows terminal startup output

* test: align reported terminal E2E contract

* fix: bind terminal browser ownership

* fix: stabilize terminal browser attachments

* fix: retry empty Windows terminal reads

* fix: preserve terminal sessions after creation

* fix: preserve terminal clipboard shortcuts

* test: align daemon harness profile resolution

* fix: harden terminal process shutdown

* fix: harden terminal approvals and agent guidance

* fix: harden terminal runtime lifecycle

* fix: unify terminal transport contracts

* fix: complete integrated terminal lifecycle

* build: regenerate integrated terminal contracts

* feat: rebuild integrated terminal window

Co-authored-by: Cursor <[email protected]>

* feat: rebuild integrated terminal journal

Co-authored-by: Cursor <[email protected]>

* feat: rebuild terminal input attention

Co-authored-by: Cursor <[email protected]>

* feat: rebuild terminal approval surfaces

Co-authored-by: Cursor <[email protected]>

* feat: rebuild session terminal blocks

Co-authored-by: Cursor <[email protected]>

* feat: rebuild terminal quote and settings

Co-authored-by: Cursor <[email protected]>

* docs: clarify estimated terminal journal boundaries

Co-authored-by: Cursor <[email protected]>

* fix: close integrated terminal host gaps

Co-authored-by: Cursor <[email protected]>

* fix: complete integrated terminal remediation

* fix: final review

* fix: close integrated terminal QA gaps

* fix: repair integrated terminal CI

* fix: stabilize integrated terminal CI

* fix: close integrated terminal web CI gaps

* fix: stabilize terminal detach chord

* test: stabilize loop event resume E2E

* ci: parallelize e2e lanes

* fix: close final web e2e races

* fix: close exact-head web e2e races

* fix: close exact-head e2e races

* test: stabilize terminal e2e fixtures

* fix: preserve semantic identity during retarget

* test: stabilize loop read E2E

* fix: stabilize integrated terminal and remove in-app tabs

Close is idempotent (exited terminals answer their recorded exit), the
stream client settles as closed instead of reconnecting a dead terminal,
size votes made before the socket opened are delivered on open, and the
fish shim rides vendor_conf.d without hiding the user's config. The web
terminal is window-native: one terminal per OS window, an id-less route
that adopts or creates, head-level Journal/New/Close verbs, and window
close never signals the PTY.

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: repair zsh markers and resolve terminals from fresh catalog truth

The zsh marker hooks shadowed zsh's read-only status parameter, killing the
precmd hook on every prompt and dropping every finish marker; a live zsh test
now pins the clean hook. The id-less terminal route waits for a catalog fetch
from this mount before adopting or creating, so a terminal opened moments
before the window cannot be missed and duplicated. Browser journeys walk the
window-native flow end to end, and the QA tracker records the walked verdicts.

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: dedupe worktrees spec imports after rebase

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: split protocol client wiring and align desktop terminal flow

Co-Authored-By: Claude Fable 5 <[email protected]>

* test: assert reattach liveness on the focused restored terminal

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: terminal fonts

* fix: render agent-internal commands as plain session output

Pull ADR-012 door 3 (`_meta.terminal_output`): the daemon no longer opts
into agent-reported terminal output, and the whole reported pipeline goes
with it — ACP ingestion and suppression, the `agent_reported` origin,
contract/stream payload fields, transcript codec/replay, acpmock steps,
and the web "reported by agent" xterm blocks. Agent-internal commands
render as plain command output again; the deliberate S3 supervised block
and the ACP pipe door stay untouched.

Co-Authored-By: Claude Fable 5 <[email protected]>

* feat: materialize agent-opened terminals as desktop windows

A daemon bridge observes terminal opened events and, for agent-actor
interactive PTY terminals (terminal_open, terminal_exec visible:true),
opens a managed Terminal window in the workspace window model — on the
desktop showing the bound session, never moving focus, one-shot per
terminal, deduped by instance key, and never failing the tool call.
Pipe terminals and human actors materialize nothing.

The transcript now recognizes real agent traffic: hosted-MCP tool names
(mcp__<server>__compozy__*) normalize to their native ids for the S3
supervised block and row labels, MCP string envelopes decode so
terminal_id reaches the block, and the remaining terminal tools carry
registered labels. E2E-003 asserts the window materializes without
stealing focus.

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: steer agents to the integrated terminal and refresh its truth surfaces

Terminal tool descriptors now lead with the desktop-window capability,
the official skill documents the windows contract and the
write-wait-read loop, and the site docs replace the agent-reported
block section with the plain-output and window-materialization truth.
QA tracker updated for ADR-019: the reported-terminal scenario is
deleted, affected scenarios were reset and re-walked with a real Haiku
agent in an isolated lab, and a new materialization scenario records
the walk (docs/qa/reports/2026-09-01-terminal-rework.md).

Co-Authored-By: Claude Fable 5 <[email protected]>

* ci: retrigger PR checks for the terminal rework head

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: renumber terminal migration and settle generated artifacts after rebase

Main's loop work took 00098/00099, so the workspace-deletion-intents
migration moves to the next gap-free version (00100), atlas.sum is
rehashed for the unpublished tail, and the SDK contract chunks
regenerate to the merged truth.

Co-Authored-By: Claude Fable 5 <[email protected]>

* fix: reconcile pointer actor and ahead sentinel after second rebase

Main's new loop tests construct ActionExecutionInput with a value actor
while this branch moved the field to *task.ActorContext; the merged call
sites now take the pointer. The global ahead-schema fixtures pinned
version 100, which the renumbered terminal migration just occupied —
they use the 999 sentinel the other streams already use.

Co-Authored-By: Claude Fable 5 <[email protected]>

* docs: increase deep review limits

* fix: remediate integrated terminal review

* test: align rebased suites with main contracts

* fix: repair rebased CI regressions

* fix: skill triggering

---------

Co-authored-by: Codex <[email protected]>
Co-authored-by: Cursor <[email protected]>
Co-authored-by: Claude Fable 5 <[email protected]>

…ozy#1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).

…ozy#1

Checkpoint via cy-loop-tasks (iteration 2, phase B mode=tasks).