@gobing-ai/spur 0.3.29 → 0.3.31

package/config/rules/strict/runtime-boundaries.yaml

@@ -60,7 +60,8 @@ rules:
             - "packages/app/src/services/token-ledger-service.ts" # FD byte-window log tailing
             - "packages/app/src/services/token-ledger-watcher.ts" # node:fs watch() live watcher
             - "packages/app/src/services/project-registry.ts" # atomic projects.json persistence
-            - "packages/app/src/observability/run-output-sink.ts" # sync FD append for mid-run tail-able artifact (task 0414 R2); observe() is sync from the lifecycle relay
+            - "packages/app/src/observability/workflow-run-log-sink.ts" # sync FD append for mid-run tail-able all-in-one run log (task 0426 / feature D2); append() is sync from the observability bus
+            - "apps/cli/src/commands/workflow.ts" # FD byte-window tail of the mid-run run log for `workflow trace --follow` streaming (task 0428 / feature D2); readSync at offset over the observability sink's FDs
             - "apps/web/src/modules/discover.ts" # Vite/Astro module scanner fallback under bun test
 
             # Synchronous bootstrap & path resolution:

package/config/templates/AGENTS.md

@@ -104,14 +104,9 @@ feature status → `05`. Working-layer, audit, and satellite rules live in the p
 
 ## Design system
 
-**Conditional contract:** If repository-root `DESIGN.md` exists, read it before planning or
-implementing any change to UI, styling, interaction, accessibility, or responsive behavior. Treat
-it as the project source of truth for visual and interaction design — tokens, components, patterns,
-and UX constraints — and keep affected work consistent with it. If it is absent, continue with the
-project's established UI conventions.
-
-Root `DESIGN.md` owns UI/UX guidance; `docs/04_DESIGN.md` still owns command, config, schema, and DTO
-surface shapes under the doc map above.
+**Conditional contract:** If repository-root `DESIGN.md` exists, leverage it dynamically as the industry-standard SSOT for UI design documentation — visual language, color tokens, typography, component specs, layout, micro-animations, accessibility, and responsive patterns. Read it before planning or implementing any UI changes, and keep affected work consistent with it. If `DESIGN.md` is absent, ignore it and continue with the project's established UI conventions.
+
+**Boundary distinction:** Root `DESIGN.md` owns UI/UX design guidance; `docs/04_DESIGN.md` owns non-UI surface design by default (command signatures, flags, config schemas, DTOs, and system boundaries). When working with design teams, choose `DESIGN.md` for UI/UX visual design and `docs/04_DESIGN.md` for non-UI API/schema surfaces.
 
 ---
 

package/config/templates/docs/04_DESIGN.md

@@ -18,12 +18,15 @@ sync: [T3, T9]
 > `docs/design/<slug>.md` satellite; this index carries the surface map + pointers.
 > Edit order: satellite first, then index row — same change (T9).
 
-## UI/UX boundary
+## UI/UX boundary & DESIGN.md
 
-Repository-root `DESIGN.md` owns all UI/UX design, including visual language, design tokens,
-components, layout, interaction, accessibility, and responsive behavior. Read and update it for UI
-work; keep this document focused on non-UI surface design. If `DESIGN.md` is absent, follow the
-project's established UI conventions rather than adding UI guidance here.
+Repository-root `DESIGN.md` owns all UI/UX design documentation (industry standard visual language, color tokens, typography, component specs, accessibility, and responsive patterns). Read and update it for UI work; keep `docs/04_DESIGN.md` focused on non-UI surface design by default. If `DESIGN.md` is absent, ignore it and follow the project's established UI conventions.
+
+By contrast, `docs/04_DESIGN.md` is our SSOT of non-UI surface design by default — covering CLI command signatures, flags, config schemas, DTOs, tables, and system boundaries.
+
+When collaborating with the design team:
+- **UI/UX & Visual Design:** Refer to and update repository-root `DESIGN.md`.
+- **Non-UI Surface & API/Schema DTOs:** Refer to and update `docs/04_DESIGN.md` (and `docs/design/<slug>.md` satellites).
 
 ## 1. CLI commands
 

package/config/templates/docs/99_PROJECT_CONSTITUTION.md

@@ -80,7 +80,8 @@ Each project's `AGENTS.md` embeds an instantiated copy of this table (§4.4). A
 | `docs/01_PRD.md` | **WHAT** — product vision, users, scope (in / out / deferred) | **Authoritative on scope** | Read before adding a command/feature; edit when scope changes |
 | `docs/02_ROADMAP.md` | **WHEN** — phases, current vs deferred, sequencing | Derived | Read to place work in a phase; edit when phase status changes |
 | `docs/03_ARCHITECTURE.md` | **HOW** — module boundaries, data flow, runtime model, invariants, rationale-in-depth | Derived (ADR wins) | Read before cross-module/seam/schema work; edit when boundaries or mechanisms change |
-| `docs/04_DESIGN.md` | **SURFACE** — concrete shapes: every CLI command, flag, config key, env var, table, DTO; **index over `docs/design/<slug>.md`** (§4.5) | Derived | Read/edit when changing a command, flag, env var, or schema — same commit |
+| `docs/04_DESIGN.md` | **SURFACE** — concrete shapes: every CLI command, flag, config key, env var, table, DTO; **index over `docs/design/<slug>.md`** (§4.5) | Derived | Read/edit when changing a non-UI command, flag, env var, or schema — same commit |
+| `DESIGN.md` (repo root) | **UI/UX SURFACE** — visual design, color tokens, typography, component specs, layout, micro-animations, accessibility | **Authoritative for UI/UX when present** | Read/edit when planning or implementing UI/UX visual changes (dynamically supported; ignored when absent) |
 | `docs/05_FEATURES.md` | **STATUS** — feature decomposition + state (✅ done / 🔶 partial / ⏳ planned / 💤 deferred); **index over `docs/features/<id>_<slug>.md`** (§4.5) | Derived | Read to find a feature's state; edit when a feature's status changes |
 | `docs/99_PROJECT_CONSTITUTION.md` | **PROCESS** — how the files above are maintained | **Authoritative on process** | Read before editing any doc above; edit per §6.8 |
 | `AGENTS.md` (repo root) | **ENTRY** — how agents work in this repo: stack, commands, gates, conventions + the instantiated doc map | Derived (from 99 + 00/01/04) | Read first every session; regenerate factual blocks from code (§6.7) |
@@ -89,7 +90,8 @@ Each project's `AGENTS.md` embeds an instantiated copy of this table (§4.4). A
 
 - Decision + one-line reason → `00`. Rationale/mechanism in depth → `03`.
 - Scope (in/out/deferred) → `01`. Mechanism / data flow / invariants → `03`.
-- Command/flag/config/schema/DTO shapes → `04`. Phase timing → `02`. Feature status → `05`.
+- UI/UX visual design, design tokens, component specs & accessibility → `DESIGN.md` (when present; otherwise follow established project UI conventions).
+- Non-UI command/flag/config/schema/DTO shapes → `04`. Phase timing → `02`. Feature status → `05`.
 - If you are writing *how it's built* or *why* inside `00`/`01`/`02`, it belongs in `03`/`04`.
 
 ### 4.2 Working layers (outside the authority chain)

package/config/workflows/basic.yaml

@@ -2,58 +2,136 @@
 # iteration bound is exhausted. Authored for the @gobing-ai/ts-dual-workflow-engine
 # state-machine schema (initialState / states[].id / onEnter / top-level transitions).
 # See Architecture Section 5 and Design Section 3.3.
+#
+# Reliability (aligned with task-pipeline / ADR-043):
+#   - check is a soft probe (status file + always exit 0) so red gates route to fix
+#     via transitions, not a raw lifecycle abort that skips the failed terminal
+#   - fix is pure slash /sp:dev-fixall against the same qualityGateCmd (single-sourced)
+#   - qualityGateMaxFixAttempts bounds the fix loop; exhausted → failed (fail-closed)
+#   - Bun package scripts only for gate surface (no alternate package-manager CLIs)
+
 "$schema": "@gobing-ai/spur/schemas/state-machine-workflow.schema.json"
 name: basic
 kind: state-machine
 description: The canonical implement-check-fix-until-pass loop
-iterationBound: 2
+iterationBound: 8
 initialState: implement
 terminalStates:
   - done
   - failed
+failureStates:
+  - failed
+vars:
+  # `task` is reserved by the engine runtime — use taskLabel for the free-form label.
+  taskLabel: "task"
+  agent: "omp"
+  stepTimeoutMs: "1800000"
+  # Override per project: `--vars '{"qualityGateCmd":"bun run lint && bun run test"}'`
+  qualityGateCmd: "bun run check"
+  # Max /sp:dev-fixall hops after a red gate (default 2). Counter: .spur/run/${vars.__runId}-basic-fix-attempt.
+  qualityGateMaxFixAttempts: "2"
+  # Injected by WorkflowAppService.run(); scopes .spur/run artifacts per run (0425 R4).
+  __runId: ""
+
 states:
   - id: implement
-    description: Implement the requested task
+    description: Implement the requested task (note-only stub; real work uses task-pipeline)
     onEnter:
       - kind: note
         options:
-          message: 'Implementing task: ${task}'
+          message: 'Implementing task: ${vars.taskLabel}'
+      - kind: shell
+        options:
+          command: >-
+            mkdir -p .spur/run &&
+            printf '0\n' > .spur/run/$__runId-basic-fix-attempt
 
   - id: check
-    description: Verify the implementation passes all checks
+    description: >
+      Soft quality-gate probe. Runs qualityGateCmd, records PASS|FAIL under
+      .spur/run/${vars.__runId}-basic-gate.status, always exit 0 so transitions branch cleanly.
     onEnter:
       - kind: shell
         options:
-          command: bun run check
+          command: >-
+            mkdir -p .spur/run &&
+            STATUS_FILE=".spur/run/$__runId-basic-gate.status" &&
+            set +e &&
+            ( $qualityGateCmd ); gate_rc=$?; set -e &&
+            if [ "$gate_rc" -eq 0 ]; then
+              printf 'PASS\n' > "$STATUS_FILE";
+            else
+              printf 'FAIL\n' > "$STATUS_FILE";
+            fi &&
+            exit 0
 
   - id: fix
-    description: Address failures from the check pass
+    description: >
+      Bounded auto-fix via pure slash /sp:dev-fixall (ADR-043) against qualityGateCmd.
     onEnter:
-      - kind: note
+      - kind: shell
         options:
-          message: 'Please fix the issues found'
+          command: >-
+            mkdir -p .spur/run &&
+            ATTEMPT_FILE=".spur/run/$__runId-basic-fix-attempt" &&
+            n=$(cat "$ATTEMPT_FILE" 2>/dev/null || echo 0) &&
+            printf '%s\n' "$((n + 1))" > "$ATTEMPT_FILE"
+      - kind: agent.run
+        options:
+          agent: ${vars.agent}
+          input: /sp:dev-fixall "${vars.qualityGateCmd}"
+          timeoutMs: ${vars.stepTimeoutMs}
 
   - id: done
     description: Terminal — workflow completed successfully
   - id: failed
-    description: Terminal — workflow failed
+    description: Terminal — quality gate still red after max fix attempts, or unrecoverable
 
 transitions:
   - from: implement
     to: check
     description: Hand the implemented task to the check pass
+    guard:
+      kind: always
 
-  # Check passed -> done; otherwise loop back to fix. Declaration order matters:
-  # the action-ok guard is tried first, so a passing check short-circuits to done.
+  # Soft probe branching: PASS first, FAIL → fix under cap, exhausted → failed.
   - from: check
     to: done
     description: Check passed — finish successfully
     guard:
-      kind: action-ok
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$__runId-basic-gate.status 2>/dev/null)" = PASS'
   - from: check
     to: fix
-    description: Check failed — route to a fix attempt
+    description: Check failed and under max fix attempts — route to fixall
+    guard:
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$__runId-basic-gate.status 2>/dev/null)" = FAIL && test "$(cat .spur/run/$__runId-basic-fix-attempt 2>/dev/null || echo 0)" -lt "$qualityGateMaxFixAttempts"'
+  - from: check
+    to: failed
+    description: Check still red after max fix attempts — stop at failed
+    guard:
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$__runId-basic-gate.status 2>/dev/null)" = FAIL && test "$(cat .spur/run/$__runId-basic-fix-attempt 2>/dev/null || echo 0)" -ge "$qualityGateMaxFixAttempts"'
+  # Defense: missing/corrupt status — attempt fix if under cap, else failed.
+  - from: check
+    to: fix
+    description: Probe status missing/corrupt and under max attempts — attempt fixall
+    guard:
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$__runId-basic-fix-attempt 2>/dev/null || echo 0)" -lt "$qualityGateMaxFixAttempts"'
+  - from: check
+    to: failed
+    description: Probe status missing/corrupt and attempts exhausted — failed
+    guard:
+      kind: always
 
   - from: fix
     to: check
-    description: Re-run the check after fixing
+    description: Re-run the soft check after fixing
+    guard:
+      kind: always

package/config/workflows/docs-pipeline.yaml

@@ -1,46 +1,79 @@
 # Docs-only task pipeline — one run per documentation/design-doc deliverable.
 #
-# This sibling of task-pipeline.yaml handles tasks whose output is a document, not a
-# code change. It keeps docs deliverables out of the implement/test/verify code gate
-# and routes them through an explicit docs-review HITL gate.
+# Sibling of task-pipeline.yaml for tasks whose output is a document, not product
+# code. Keeps docs deliverables out of the full quality gate and routes them
+# through an explicit docs-review HITL gate.
+#
+# Reliability (aligned with task-pipeline / ADR-043):
+#   - Soft precheck (task check + status file) so red precheck → failed via transitions
+#   - draft agent.run is a pure slash command (docs discipline lives in
+#     /sp:dev-run --mode implement + sp:code-implementation, not YAML prose)
+#   - requireDiff fails silent no-op drafts
+#   - HITL answer routing is exhaustive (yes/no/cancel)
+#   - record → done is fail-closed on `spur task check`
 
 "$schema": "@gobing-ai/spur/schemas/state-machine-workflow.schema.json"
 kind: state-machine
 name: docs-pipeline
-description: "Docs-only execution pipeline (draft -> docs-review -> record -> done)"
+description: "Docs-only execution pipeline (precheck -> draft -> docs-review -> record -> done)"
 iterationBound: 12
-initialState: draft
+initialState: precheck
 terminalStates:
   - done
   - failed
   - cancelled
+failureStates:
+  - failed
+  - cancelled
 vars:
   wbs: "0000"
   profile: "standard"
   spurBin: "spur"
   agent: "omp"
-  stepTimeoutMs: "600000"
+  stepTimeoutMs: "1800000"
   __hitlAnswer: ""
 
 states:
+  - id: precheck
+    description: >
+      Soft precheck: task check + agent doctor. Status files always exit 0 so
+      failures route to `failed` via transitions (not a raw lifecycle abort).
+    onEnter:
+      - kind: shell
+        options:
+          command: >-
+            mkdir -p .spur/run &&
+            set +e &&
+            $spurBin task check $wbs; check_rc=$?;
+            $spurBin agent doctor $agent; doctor_rc=$?;
+            set -e &&
+            if [ "$check_rc" -eq 0 ] && [ "$doctor_rc" -eq 0 ]; then
+              printf 'PASS\n' > ".spur/run/$wbs-docs-precheck.status";
+            else
+              printf 'FAIL\n' > ".spur/run/$wbs-docs-precheck.status";
+            fi &&
+            exit 0
+      - kind: note
+        options:
+          message: "Docs pipeline start for task ${vars.wbs}."
+
   - id: draft
     description: >
-      Author or update the documentation deliverable for task ${vars.wbs}. This is a
-      docs-only path: do not run implementation, unit-test generation, or code review
-      steps from task-pipeline.yaml.
+      Author or update the documentation deliverable for task ${vars.wbs}. Pure slash
+      implement hop (ADR-043). Docs-only scope is owned by the implement competency when
+      the task template/background is documentation — not by YAML essays.
     onEnter:
       - kind: agent.run
         options:
           agent: ${vars.agent}
-          input: >-
-            /sp:dev-run --mode implement ${vars.wbs} --auto.
-            Treat this as a docs-only/design-doc deliverable: update the requested
-            documentation and do not add product code or tests unless the task
-            explicitly requires them.
+          # Pure slash only (ADR-043). Do not bolt docs-only prose onto the input line.
+          input: /sp:dev-run --mode implement ${vars.wbs} --auto
           timeoutMs: ${vars.stepTimeoutMs}
+          # Fail a silent no-op draft (same empty-implement class as task-pipeline R3).
+          requireDiff: true
       - kind: shell
         options:
-          command: "${vars.spurBin} task update ${vars.wbs} wip --no-lifecycle"
+          command: "$spurBin task update $wbs wip --no-lifecycle"
 
   - id: docs-review
     description: >
@@ -60,7 +93,7 @@ states:
     onEnter:
       - kind: shell
         options:
-          command: "${vars.spurBin} task record ${vars.wbs} --solution-from-diff --transition testing"
+          command: "$spurBin task record $wbs --solution-from-diff --transition testing"
 
   - id: done
     description: Terminal — docs deliverable reviewed and recorded.
@@ -72,11 +105,11 @@ states:
         options:
           command: >
             mkdir -p .spur/run &&
-            printf '%s\n' "{\"wbs\":\"${vars.wbs}\",\"verdict\":\"PASS\",\"source\":\"docs-pipeline\",\"requirements\":[{\"id\":\"docs-deliverable\",\"status\":\"MET\",\"evidence\":\"docs-pipeline record step\"}],\"acceptanceCriteria\":[],\"checks\":[{\"name\":\"docs-only\",\"status\":\"pass\",\"evidence\":\"docs-pipeline\"}]}"
-            > ".spur/run/${vars.wbs}-verdict.json"
+            printf '%s\n' "{\"wbs\":\"$wbs\",\"verdict\":\"PASS\",\"source\":\"docs-pipeline\",\"requirements\":[{\"id\":\"docs-deliverable\",\"status\":\"MET\",\"evidence\":\"docs-pipeline record step\"}],\"acceptanceCriteria\":[],\"checks\":[{\"name\":\"docs-only\",\"status\":\"pass\",\"evidence\":\"docs-pipeline\"}]}"
+            > ".spur/run/$wbs-verdict.json"
       - kind: shell
         options:
-          command: "${vars.spurBin} task update ${vars.wbs} done --no-lifecycle"
+          command: "$spurBin task update $wbs done --no-lifecycle"
       - kind: note
         options:
           message: "Docs pipeline complete for task ${vars.wbs}."
@@ -88,13 +121,26 @@ states:
     description: Terminal — docs pipeline cancelled by operator.
 
 transitions:
+  - from: precheck
+    to: draft
+    description: Precheck PASS — begin docs draft.
+    guard:
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$wbs-docs-precheck.status 2>/dev/null)" = PASS'
+  - from: precheck
+    to: failed
+    description: Precheck FAIL — stop before draft.
+    guard:
+      kind: always
+
   - from: draft
     to: record
     description: profile=auto — skip the docs-review HITL gate.
     guard:
       kind: shell
       options:
-        command: 'test "${vars.profile}" = auto'
+        command: 'test "$profile" = auto'
   - from: draft
     to: docs-review
     description: Interactive — enter docs-review.
@@ -107,21 +153,21 @@ transitions:
     guard:
       kind: shell
       options:
-        command: 'test "${vars.__hitlAnswer}" = yes'
+        command: 'test "$__hitlAnswer" = yes'
   - from: docs-review
     to: failed
     description: Docs rejected — stop for revision.
     guard:
       kind: shell
       options:
-        command: 'test "${vars.__hitlAnswer}" = no'
+        command: 'test "$__hitlAnswer" = no'
   - from: docs-review
     to: cancelled
     description: Operator cancelled docs review.
     guard:
       kind: shell
       options:
-        command: 'test "${vars.__hitlAnswer}" = cancel'
+        command: 'test "$__hitlAnswer" = cancel'
 
   - from: record
     to: done
@@ -129,11 +175,11 @@ transitions:
     guard:
       kind: shell
       options:
-        command: "${vars.spurBin} task check ${vars.wbs}"
+        command: "$spurBin task check $wbs"
   - from: record
     to: failed
     description: Task check failed — block before done.
     guard:
       kind: shell
       options:
-        command: '! ${vars.spurBin} task check ${vars.wbs}'
+        command: '! $spurBin task check $wbs'

package/config/workflows/feature-dev.yaml

@@ -1,11 +1,19 @@
+# Feature development loop — brainstorm → plan → execute-tasks → feature-verify → done.
+#
+# Reliability (aligned with task-pipeline / ADR-043):
+#   - Soft precheck (featureId non-empty + agent doctor) → failed via transitions
+#   - Every agent.run input is a pure slash command (no free-form pipeline essays)
+#   - featureId must be supplied by the caller (workflow vars are not agent-mutated)
+#   - feature-verify is fail-closed on `spur feature check --strict`
+
 "$schema": "@gobing-ai/spur/schemas/state-machine-workflow.schema.json"
 kind: state-machine
 name: feature-dev
 description: >
-  End-to-end FEATURE development loop: brainstorm → plan (decompose) → execute-tasks →
-  feature-verify → done. Drives a whole feature (one description → a feature with BDD AC →
-  many tasks → all tasks done → feature verified), NOT a single task — for one deliverable
-  use task-pipeline.yaml directly.
+  End-to-end FEATURE development loop: precheck → brainstorm → plan (decompose) →
+  execute-tasks → feature-verify → done. Drives a whole feature (one description → a
+  feature with BDD AC → many tasks → all tasks done → feature verified), NOT a single
+  task — for one deliverable use task-pipeline.yaml directly.
 
   The engine has no native for-each over a dynamic task list, so the execute-tasks step
   DELEGATES the fan-out: a single agent.run instructs the agent to run every pending task
@@ -13,70 +21,106 @@ description: >
   This mirrors how /sp:dev-run --mode full delegates its inner work loop (ADR-022: orchestration
   is configuration — zero new engine code).
 
-  Shape: brainstorm → plan → execute-tasks(delegated loop) → feature-verify → done
-         (plan failure or a non-PASS task short-circuits to `failed`).
+  Shape: precheck → brainstorm → plan → execute-tasks[-auto](delegated loop) → feature-verify → done
+         (precheck/plan failure or a non-PASS feature check short-circuits to `failed`).
 
   Vars (passed as a JSON object via `--vars`):
     featureId — the feature ID this loop drives (required), e.g. `--vars '{"featureId":"B3"}'`
     agent     — agent the agent.run steps invoke; override `--vars '{"agent":"claude"}'`
-    profile   — `auto` skips the per-task HITL approval inside task-pipeline
+    profile   — `auto` selects execute-tasks-auto (`/sp:dev-runall … --auto`); otherwise
+                interactive execute-tasks (no --auto, per-task HITL remains)
     spurBin   — PATH-independent spur invocation (CLI overrides at run start; literal fallback)
+    stepTimeoutMs — agent.run budget (ms) for brainstorm/plan/runall
 
-  Seeded by `spur init`; adapt the agent.run inputs to your project's command set.
+  Reliability (ADR-043): every agent.run input is a pure slash command. featureId must be
+  supplied by the caller (workflow vars cannot be mutated by the agent). profile is wired
+  via two execute hops (templates cannot branch a single input string).
 iterationBound: 20
-initialState: brainstorm
+initialState: precheck
 terminalStates:
   - done
   - failed
+failureStates:
+  - failed
 vars:
   featureId: ""
   agent: "omp"
   profile: "standard"
   spurBin: "spur"
+  stepTimeoutMs: "1800000"
+  # Injected by WorkflowAppService.run(); scopes .spur/run artifacts per run (0425 R4).
+  __runId: ""
 
 states:
+  - id: precheck
+    description: >
+      Soft precheck: non-empty featureId + agent doctor. Always exit 0 so failures
+      route to `failed` via transitions (not a raw lifecycle abort).
+    onEnter:
+      - kind: shell
+        options:
+          command: >-
+            mkdir -p .spur/run &&
+            set +e &&
+            test -n "$featureId"; id_rc=$?;
+            $spurBin agent doctor $agent; doctor_rc=$?;
+            set -e &&
+            if [ "$id_rc" -eq 0 ] && [ "$doctor_rc" -eq 0 ]; then
+              printf 'PASS\n' > ".spur/run/$__runId-feature-dev-precheck.status";
+            else
+              printf 'FAIL\n' > ".spur/run/$__runId-feature-dev-precheck.status";
+            fi &&
+            exit 0
+      - kind: note
+        options:
+          message: "Feature-dev start for feature ${vars.featureId}."
+
   - id: brainstorm
     description: >
-      Turn vague intent into a validated feature with BDD acceptance criteria via
-      /sp:dev-brainstorm --feature. The grilling interview maps the decision space; the
-      --feature exit authors AC and loops `spur feature check` until clean. Record the
-      derived feature ID into ${vars.featureId} for the downstream steps.
+      Validate / expand the feature with BDD AC via pure slash /sp:dev-brainstorm.
+      Caller must pass vars.featureId (required for plan/runall).
     onEnter:
       - kind: agent.run
         options:
           agent: ${vars.agent}
-          input: /sp:dev-brainstorm --feature
+          input: /sp:dev-brainstorm --feature ${vars.featureId}
+          timeoutMs: ${vars.stepTimeoutMs}
 
   - id: plan
     description: >
-      Decompose the validated feature into a CLI-validated task batch via /sp:dev-plan.
-      The batch-create gate (task-batch.schema.json) rejects a malformed decomposition,
-      so a bad plan never lands tasks.
+      Decompose the validated feature into a CLI-validated task batch via pure slash
+      /sp:dev-plan. The batch-create gate rejects a malformed decomposition.
     onEnter:
       - kind: agent.run
         options:
           agent: ${vars.agent}
           input: /sp:dev-plan --feature ${vars.featureId}
+          timeoutMs: ${vars.stepTimeoutMs}
+
+  # Two pure-slash execute hops so vars.profile actually controls --auto (ADR-043).
+  # Engine templates cannot branch a single input string on profile; declaration-order
+  # transitions pick the correct hop (auto first, then interactive fallback).
+  - id: execute-tasks-auto
+    description: >
+      profile=auto — drive every pending task via pure slash /sp:dev-runall --auto
+      (skips per-task HITL inside task-pipeline).
+    onEnter:
+      - kind: agent.run
+        options:
+          agent: ${vars.agent}
+          input: /sp:dev-runall --feature ${vars.featureId} --auto
+          timeoutMs: ${vars.stepTimeoutMs}
 
   - id: execute-tasks
     description: >
-      Drive EVERY pending task under the feature to done (e.g. `/sp:dev-runall --feature ${featureId} --auto`).
-      The agent enumerates the work with `spur task list --feature ${featureId} --status todo --json`,
-      then runs each task through task-pipeline.yaml (precheck → implement → test → review → approve → verify → record → done).
-      It stops and reports on the first non-PASS verdict — a failing task fails the feature.
+      Interactive — drive every pending task via pure slash /sp:dev-runall without
+      --auto (per-task HITL approve gates remain available).
     onEnter:
       - kind: agent.run
         options:
           agent: ${vars.agent}
-          input: >
-            Run every pending task under feature ${vars.featureId} to done. Enumerate with
-            `${vars.spurBin} task list --feature ${vars.featureId} --status todo --json`; then
-            for each WBS run the standard single-task pipeline VERBATIM:
-            `${vars.spurBin} workflow run .spur/workflows/task-pipeline.yaml --vars
-            '{"wbs":"<wbs>","profile":"${vars.profile}","agent":"${vars.agent}"}' --async --json`,
-            polling `${vars.spurBin} workflow trace <run-id>` until terminal and inspecting
-            `.spur/run/<wbs>-verdict.json`. Stop and report on the first non-PASS verdict
-            (a failing task fails the feature).
+          input: /sp:dev-runall --feature ${vars.featureId}
+          timeoutMs: ${vars.stepTimeoutMs}
 
   - id: feature-verify
     description: >
@@ -90,12 +134,25 @@ states:
     onEnter:
       - kind: shell
         options:
-          command: 'mkdir -p .spur/memory/sessions && echo "checkpoint: feature-dev done featureId=${vars.featureId} ts=$(date -u +%Y-%m-%dT%H:%M:%SZ)" > .spur/memory/sessions/${vars.featureId}-checkpoint.md'
+          command: 'mkdir -p .spur/memory/sessions && echo "checkpoint: feature-dev done featureId=$featureId ts=$(date -u +%Y-%m-%dT%H:%M:%SZ)" > .spur/memory/sessions/$featureId-checkpoint.md'
 
   - id: failed
     description: Terminal — plan rejected, a task failed verification, or the feature check did not pass.
 
 transitions:
+  - from: precheck
+    to: brainstorm
+    description: Precheck PASS — begin brainstorm.
+    guard:
+      kind: shell
+      options:
+        command: 'test "$(cat .spur/run/$__runId-feature-dev-precheck.status 2>/dev/null)" = PASS'
+  - from: precheck
+    to: failed
+    description: Precheck FAIL (missing featureId and/or doctor red) — stop.
+    guard:
+      kind: always
+
   # Linear prefix: brainstorm → plan → execute-tasks
   - from: brainstorm
     to: plan
@@ -103,15 +160,27 @@ transitions:
     guard:
       kind: always
 
+  - from: plan
+    to: execute-tasks-auto
+    description: Tasks created; profile=auto — batch with --auto (no per-task HITL).
+    guard:
+      kind: shell
+      options:
+        command: 'test "$profile" = auto'
   - from: plan
     to: execute-tasks
-    description: Tasks created; execute them
+    description: Tasks created; interactive — batch without --auto.
     guard:
       kind: always
 
+  - from: execute-tasks-auto
+    to: feature-verify
+    description: Auto batch finished — verify the feature.
+    guard:
+      kind: always
   - from: execute-tasks
     to: feature-verify
-    description: All tasks reported done; verify the feature
+    description: Interactive batch finished — verify the feature.
     guard:
       kind: always
 
@@ -123,11 +192,11 @@ transitions:
     guard:
       kind: shell
       options:
-        command: "${vars.spurBin} feature check ${vars.featureId} --strict"
+        command: "$spurBin feature check $featureId --strict"
   - from: feature-verify
     to: failed
     description: Strict feature check failed — block before done; investigate incomplete tasks or AC.
     guard:
       kind: shell
       options:
-        command: '! ${vars.spurBin} feature check ${vars.featureId} --strict'
+        command: '! $spurBin feature check $featureId --strict'