* fix(gateway): snapshot agentctl status and env id on session subscribe
Late-subscribing clients (page reload, task switch, WS reconnect) miss
the agentctl_ready and task_environment_id events fired during the
original launch. Without them, the env-routed shell terminal gate stays
stuck and the panel hangs forever on "Connecting terminal...".
- appendSessionStateMessage now carries task_environment_id so the
frontend can seed environmentIdBySessionId for the active session.
- New appendAgentctlStatusMessage snapshots current agentctl readiness
via a bounded Health probe and emits session.agentctl_ready /
_starting to flip the client gate out of `undefined`.
- GetOrEnsureExecutionForEnvironment now publishes AgentctlStarting to
mirror EnsureWorkspaceExecutionForSession and keep gate transitions
monotonic.
* fix(web): drop shell-mode agentctl gate to unblock env-routed terminals
The shell terminal gate required agentctlStatus.isReady to flip to ready
before opening the WS, but that event may have fired before the client
subscribed (page reload, task switch, local-executor task switch). The
lazy create-on-terminal-connect path on the backend never fires either,
because the WS never opens — chicken-egg.
Backend env handler already waits for remote workspace readiness
server-side (waitForRemoteExecutionReadyWithTimeout), so the client
just needs the routing keys (env id + session id). Agent-mode terminals
still gate on readiness — the agent process must be wired before
passthrough I/O is meaningful.
Extract computeCanConnect helper to keep PassthroughTerminal under the
complexity ceiling.
* test: cover terminal-hang on subscribe snapshot and env-routed gating
Adds a backend unit test asserting appendSessionStateMessage carries
task_environment_id (the deterministic RED for the env-map seed bug),
plus E2E scenarios for the user-facing repros: cold load, kanban switch,
hard reload, sidebar switch, and a local-executor task switch that used
to poison the terminal for every subsequent task. Adds an
expectTerminalConnected helper to SessionPage to detect the
"Connecting terminal..." overlay.
* fix(repo): heal task_environments on boot and enforce one env per task
Live DB audit on the user's machine surfaced two corrupt-data classes that
manifest as the user-facing "terminal doesn't appear / Failed to fetch
terminals: 503" symptoms:
- 15 worktree-mode envs with worktree_path set but workspace_path empty.
GetOrEnsureExecutionForEnvironment returns ErrSessionWorkspaceNotReady
forever, the env terminal handler 503s.
- 3 tasks with two task_environments rows (race in lazy create — pairs
4-6 s apart, both ready, distinct workspace_path).
healTaskEnvironmentWorkspacePaths backfills workspace_path = worktree_path
on the affected rows, and healDuplicateTaskEnvironments collapses each
duplicated task to its most recently updated env (re-pointing any sessions
still on the loser, then dropping the loser row).
ensureTaskEnvironmentTaskUniqueIndex adds a UNIQUE index on
task_environments(task_id) so a future race fails loud. Order matters:
heal must run first to drop duplicates before the unique index is created.
* fix(repo): reject empty workspace_path on worktree envs
The orchestrator's persistTaskEnvironment update branch previously left
workspace_path untouched, so an env created with empty workspace_path
would stay corrupt forever — terminals 503 with ErrSessionWorkspaceNotReady.
Tighten the repository boundary: CreateTaskEnvironment refuses worktree-mode
inserts without workspace_path, and UpdateTaskEnvironment refuses to clear
it back to empty. A future writer regression now fails loud at the boundary
instead of producing a row that breaks the env terminal handler.
* fix(executor): refresh paths on env update and dedupe by task
The persistTaskEnvironment update branch only mutated AgentExecutionID +
Status, so envs created earlier with an empty workspace_path stayed
permanently broken across launches. Refresh worktree + workspace fields
on update via a shared computeWorkspacePath helper used by both branches.
Concurrent launches for the same task previously raced inside this
function — each saw existingEnv == nil and each called
CreateTaskEnvironment, producing duplicate rows ~4-6 s apart in the live
DB. Hold a per-task mutex (mirroring sessionLocks) and re-fetch the env
inside the critical section so siblings reuse the row the first launch
persisted. The repo's unique index on task_environments(task_id) is the
durable guarantee; the lock just closes the window before the constraint
trips.
* test(repo): cover task_environments heal, validation, unique index
Repository unit tests for the new heal steps and writer guards:
- HealTaskEnvironmentWorkspacePaths_BackfillsEmpty / LeavesPopulatedAlone /
Idempotent
- HealDuplicateTaskEnvironments_KeepsMostRecent (asserts session re-link to
winner) / NoOpWhenSingle
- EnsureTaskEnvironmentTaskUniqueIndex_BlocksFutureDuplicates
- CreateTaskEnvironment_RejectsEmptyWorkspaceForWorktree /
AllowsNonWorktreeWithEmptyWorkspace
- UpdateTaskEnvironment_RejectsClearingWorkspaceForWorktree
Tests live in the sqlite package so they can exercise the unexported
heal* helpers directly, mirroring how the boot path calls them.
* fix: address PR review feedback
- helpers.go: drop the 2 s blocking Health probe in
appendAgentctlStatusMessage. sendSessionData runs in the WS read loop, so
the probe was delaying every subscribe/focus ACK by up to 2 s. Use
GetWorkspaceStream() != nil as the cached readiness signal — the stream
manager only attaches it after Health succeeds, so its presence is a
free, non-blocking proxy. Subsequent waitForAgentctlReady events still
correct the snapshot if it picks the wrong one.
- manager_execution.go: publish AgentctlStarting INSIDE createExecution,
before spawning waitForAgentctlReady. The previous order published it
after createExecution returned, which let the goroutine race past on
fast Health and deliver Ready before Starting. Drop the now-redundant
publishes in EnsureWorkspaceExecutionForSession and the env lazy-create
path.
- task_environment.go: replace the local executorTypeWorktree constant
with models.ExecutorTypeWorktree (canonical) so a future rename of the
type is caught by the compiler.
- base.go: heal_workspace_path now respects the kandev path layout —
task-dir-mode envs (.../.kandev/tasks/<name>/<repo>) get
Dir(worktree_path) as workspace_path, mirroring computeWorkspacePath
in the orchestrator. Plain worktree mode unchanged.
- heal_test.go: tighten unique-constraint assertion to require a UNIQUE
error string so the test fails loudly if a different DB error masks
the constraint violation. Add task-dir-mode heal coverage.
* feat(backend): key user shells by TaskEnvironmentID
Sessions in the same task share a TaskEnvironment. Re-key the user shell
runner so terminals are scoped per env, not per session — switching between
sessions of the same task now sees the same shell list.
- Add TaskEnvironmentID to AgentExecution / LaunchRequest / ExecutorCreateRequest
/ executor.LaunchAgentRequest, propagated through ToAgentExecution and the
orchestrator launch path.
- Add Manager.ResolveScopeKey(sessionID) — returns env when execution carries
it, falls back to sessionID for legacy paths.
- Rename runner param sessionID -> scopeID in interactive_shells.go; map keys
are now scopeID:terminalID.
- Shell handlers + WS terminal handler resolve sid -> scopeID via
ResolveScopeKey before calling the runner.
* feat(web): key dockview layout + portals to TaskEnvironmentID
Switching between sessions of the same task no longer rebuilds the dockview
layout — terminals and panels persist because layouts are env-scoped, not
session-scoped. The bottom terminal also reuses a stable sessionId across
same-env switches so it doesn't reconnect to a different shell.
- Rename layout persistence: setSessionLayout/getSessionLayout/*SessionMaximizeState
-> setEnvLayout/getEnvLayout/*EnvMaximizeState (kandev.dockview.env-layout. prefix).
- Rename dockview-session-switch.ts -> dockview-env-switch.ts; performSessionSwitch
-> performEnvSwitch with oldEnvId/newEnvId/activeSessionId.
- dockview-store.ts: currentLayoutSessionId -> currentLayoutEnvId; switchSessionLayout
-> switchEnvLayout(oldEnvId, newEnvId, activeSessionId). Same-env switches no-op.
Drop _skipLayoutSwitchForSession + consumeSkipFlag — env-keyed action handles it
structurally.
- panel-portal-manager: PortalEntry.envId; releaseBySession -> releaseByEnv.
ENV_SCOPED_COMPONENTS in dockview-desktop-layout (renamed from SESSION_SCOPED).
- useEnvSwitchCleanup replaces useSessionSwitchCleanup; tracks env changes only.
- All performLayoutSwitch callers (sidebar, dropdown, header, dialog) resolve
sid -> envId via environmentIdBySessionId before switching.
- bottom-terminal-panel: drop sessionId prop, use useEnvironmentSessionId().
- cleanupTaskStorage now takes envIds[] alongside sessionIds[] for env-scoped
layout/maximize cleanup.
* test: add coverage for env-keyed terminals + layout
Playwright e2e (terminal-env-keyed.spec.ts) — switching between sessions of
the same task keeps terminal scrollback intact and asserts both sessions
share one task_environment_id. Update the comment in
sessionless-task-switch.spec.ts to reference performEnvSwitch.
* fix: address env layout PR feedback
* test: fix flaky handshake auth check
* fix: address env scope review findings
* feat(backend): per-launch authentication for agentctl
Add nonce-based handshake protocol to secure all agentctl HTTP/WS
endpoints. On launch, the parent (backend) passes a one-time bootstrap
nonce via AGENTCTL_BOOTSTRAP_NONCE env var. agentctl generates its own
auth token, enables Bearer auth immediately, and returns the token to
the parent via POST /auth/handshake (validated by nonce).
Standalone, Docker, and Sprites executors all use the same handshake
flow. Removes the old AGENTCTL_AUTH_TOKEN env var approach. For
Docker/Sprites, the retrieved token is persisted encrypted in
SecretStore (AES-256-GCM) for crash recovery.
Security properties:
- Auth enabled from boot (no unprotected window)
- Token never in env vars (only the burned nonce)
- Nonce is one-shot (replay impossible)
- Constant-time token comparison (timing-safe)
* fix: check fmt.Sscanf return value in integration test
* fix: address review feedback - constant-time nonce comparison, safe ID truncation
* fix(backend): exempt MCP routes from agentctl auth middleware
The agent subprocess connects to agentctl's MCP endpoints (/sse, /message,
/mcp) to use kandev tools (create_task, ask_user_question, etc.). These
endpoints run on the same HTTP server as the authenticated API, but the
agent process doesn't possess the auth token — it was only exchanged
between the backend and agentctl during the handshake.
Exempt /sse, /message, and /mcp from the Bearer auth middleware since the
agent process runs in the same trust boundary as agentctl.
* refactor(backend): remove query param token fallback and document SetAuthToken constraint
Remove the ?token= query parameter fallback from extractBearerToken —
tokens in URLs leak via access logs, referer headers, and browser history.
All WS callers already use Authorization headers via wsAuthHeaders().
Document SetAuthToken's concurrency constraint: it must not be called
concurrently with HTTP requests on the same client.
* fix(backend): enforce mutex in ControlClient/Launcher auth, fix Start() lock scope
- Add sync.Mutex to ControlClient; protect SetAuthToken and AuthToken() reads
- Guard Launcher.AuthToken() with l.mu (written under lock in Start, read unprotected)
- Extract buildAndStartProcess and performHandshake from Start() (was 106 lines, over 80-line limit)
- Release l.mu before blocking health/handshake calls so monitorExit() can close
l.exited — previously the fast-fail path in waitForHealthy was unreachable
- Update stale executor_backend.go comment (standalone no longer uses env var auth)
Co-authored-by: João Salavisa <undefined@users.noreply.github.com>
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Co-authored-by: João Salavisa <undefined@users.noreply.github.com>