@gobing-ai/spur 0.3.89 → 0.3.91

package/.claude-plugin/marketplace.json

@@ -7,7 +7,7 @@
     "plugins": [
         {
             "name": "sp",
-            "version": "0.3.89",
+            "version": "0.3.91",
             "source": "./plugins/sp"
         }
     ]

package/config/pipeline-budgets.json

@@ -48,6 +48,13 @@
             "tokenCostUsd": null,
             "source": "Coverage entry (0826 R2): the live definition's model-query count (extractResolvedWorkflowFacts, 2026-09-11). No real-run measurement yet: wall-clock and cost are unenforced until measured, never 0.",
             "decision": null
+        },
+        "decision-routing-example": {
+            "modelQueries": 0,
+            "wallClockMs": null,
+            "tokenCostUsd": null,
+            "source": "Coverage entry (0911): authoring example (config/workflows/decision-routing-example.yaml) enumerating `decision:` modes. Measured 0 model queries — the producer is a shell action, so the example runs with no provider (live count via loadQueryCounts, 2026-09-20). Not a pipeline humans run; wall-clock and cost unenforced, never 0.",
+            "decision": null
         }
     }
 }

package/config/rules/boundary/sp-plugin-standalone.yaml

@@ -0,0 +1,37 @@
+$schema: "@gobing-ai/spur/schemas/rule-file.schema.json"
+# Plugin standalone contract — bundled plugin surfaces import only node:* or relative paths.
+#
+# WHY: `superskill install` bundles plugins/sp hooks and scripts on targets (and from
+# marketplace snapshots) with NO node_modules. A bare `@gobing-ai/*` value import only
+# converts if the host superskill happens to carry that package in its private dep tree
+# (superskill 0.3.28+ bundles ts-utils/ts-runtime; older versions do not). When it does
+# not: "Error: Bundle failed" and the plugin cannot install — this broke releases
+# 0.3.81–0.3.88 (task 0669; regression commit 84c87a9ac).
+#
+# Contract:
+# - plugins/sp/hooks/**, plugins/sp/scripts/**, plugins/sp/lib/** may value-import
+#   only `node:*` builtins and relative paths (vendored helpers live in plugins/sp/lib/).
+# - `import type` / `export type` from any package is fine — erased before bundling
+#   (e.g. inline-run-setup.ts's type-only import from @gobing-ai/app).
+# - This rule is the author-time tripwire over DIRECT imports; the transitive closure
+#   and the real `superskill install` path are checked by `bun run plugin-smoke`, and
+#   script-contract-check Rule 5 (manifest-aware) fails the spur-check gate.
+#
+# Scope: plugins/sp bundled surfaces only. apps/packages follow the normal workspace
+# import rules and are not affected.
+include:
+  - "plugins/sp/hooks/**/*.ts"
+  - "plugins/sp/scripts/**/*.ts"
+  - "plugins/sp/lib/**/*.ts"
+rules:
+  - id: sp-plugin-standalone
+    description: >
+      No value imports of workspace packages (`@gobing-ai/*` or any bare specifier)
+      in plugin bundled surfaces — superskill bundles them onto targets with no
+      node_modules. Vendor the helper into plugins/sp/lib/ (see lib/env.ts) or use a
+      relative import. `import type` is exempt.
+    severity: error
+    evaluator:
+      type: rg
+      config:
+        pattern: "^import \\{[^}]*\\} from ['\\\"]@gobing-ai/|^import ['\\\"]@gobing-ai/|^import\\(['\\\"]@gobing-ai/|^} from ['\\\"]@gobing-ai/"

package/config/workflows/decision-routing-example.yaml

@@ -0,0 +1,134 @@
+# DecisionMaker routing example (task 0911, D7). Demonstrates the three `decision:` modes
+# on hitl.* actions; copy the pattern into your own workflow instead of running this file.
+#
+#   never    — the action always asks the human (DecisionMaker is bypassed even when
+#              `workflow.hitlDecisionMaker: true`). Used for gates that must stay human-only.
+#   evidence — the DecisionMaker answers from recorded action evidence of the listed producer
+#              states: it picks one of the offered choices when confidence >= 0.9, or defers
+#              (answer var is cleared, status var records `deferred`) when evidence is missing,
+#              stale, or inconclusive. Requires a non-paused state, a statusVar distinct from
+#              the answer var, and producer states that exist in this workflow.
+#   (omitted) — legacy 0910 semantics: DecisionMaker may confirm/select without evidence and
+#              without provenance. Kept for compatibility; prefer an explicit mode.
+#
+# The run is safe and local: one shell producer emits a fixed failure summary, the choose step
+# reads it as evidence, and the operator gate has no side effects. No provider is required —
+# without `workflow.hitlDecisionMaker: true` (or without a key) the evidence step defers and
+# `never` behaves exactly like an unconfigured gate. See `spur self status` -> `decisionMaker`
+# for offline readiness. Optional registered-summary evidence (`summaryArtifact`) is documented
+# in docs/help/how_to_enable_jev_in_spur.md; this example deliberately creates no artifact.
+"$schema": "@gobing-ai/spur/schemas/state-machine-workflow.schema.json"
+kind: state-machine
+name: decision-routing-example
+version: "1"
+initialState: assess
+description: >
+  Worked example of DecisionMaker decision routing (0911 D7): a shell producer records a fixed
+  recovery summary as action evidence, a non-pausing hitl.select lets the DecisionMaker choose
+  retry or stop from that evidence (deferring with provenance when it is inconclusive), an
+  accepted retry runs exactly one attempt and then escalates, and the never-mode pause=true
+  hitl.confirm gate always waits for the operator with explicit yes/no/cancel branches.
+  Illustrative only — copy the patterns rather than running this file.
+vars:
+    recoveryChoice: ""
+    decisionStatus: ""
+    # Default answer var the never-mode gate's yes/no/cancel branches read (matches the stock gates).
+    __hitlAnswer: ""
+states:
+  - id: assess
+    description: >
+      Producer state: one shell action whose redacted output becomes the evidence the decide
+      state's DecisionMaker reads. One producer state carries one evidence-bearing action.
+    onEnter:
+      - kind: shell
+        options:
+          command: >-
+            printf 'Failure summary: verification failed twice on the flaky fixture; no data loss;
+            retry is safe, or stop and hand the decision to the operator.'
+  - id: decide
+    description: >
+      Evidence-mode select (non-paused): DecisionMaker picks a recovery step from the assess
+      state's recorded evidence. Low confidence, missing or stale evidence defers and writes the
+      status variable instead of answering, so no implicit approval is ever produced.
+    onEnter:
+      - kind: hitl.select
+        options:
+          prompt: "Choose the next safe recovery step using the recorded failure summary."
+          options: [retry, stop]
+          var: recoveryChoice
+          decision:
+            mode: evidence
+            statusVar: decisionStatus
+            evidenceNodes: [assess]
+  - id: retry
+    description: >
+      The single bounded retry an accepted `retry` choice buys: this state runs once and routes
+      to the operator gate, so an accepted answer can never loop back into the decision.
+    onEnter:
+      - kind: shell
+        options:
+          command: "printf 'retry attempt: re-running the failed verification step once'"
+  - id: humanGate
+    description: >
+      Never-mode confirm (paused): always waits for the operator, whichever branch reached it.
+      `spur workflow continue --answer yes|no|cancel` records the explicit operator answer.
+    pause: true
+    onEnter:
+      - kind: hitl.confirm
+        options:
+          prompt: "Publish the recovered result, stop, or cancel the run?"
+          decision:
+            mode: never
+  - id: published
+    description: Operator approved the recovered result; terminal marker state.
+  - id: stop
+    description: Recovery stopped deliberately — requested by the decision or rejected by the operator.
+  - id: cancelled
+    description: Operator cancelled the run at the gate; terminal marker state.
+transitions:
+  # Accepted decision branches first: the state machine takes the first passing edge.
+  - from: assess
+    to: decide
+    guard: { kind: always }
+  - from: decide
+    to: retry
+    guard:
+      kind: shell
+      options:
+        command: 'test "$recoveryChoice" = "retry"'
+  - from: decide
+    to: stop
+    guard:
+      kind: shell
+      options:
+        command: 'test "$recoveryChoice" = "stop"'
+  # Declared defer edge: evidence mode wrote `deferred` and cleared the answer var.
+  - from: decide
+    to: humanGate
+    guard:
+      kind: shell
+      options:
+        command: 'test "$decisionStatus" = "deferred"'
+  - from: retry
+    to: humanGate
+    guard: { kind: always }
+  # Explicit yes / no / cancel branches on the default answer var, matching the stock gates.
+  - from: humanGate
+    to: published
+    guard:
+      kind: shell
+      options:
+        command: 'test "$__hitlAnswer" = "yes"'
+  - from: humanGate
+    to: stop
+    guard:
+      kind: shell
+      options:
+        command: 'test "$__hitlAnswer" = "no"'
+  - from: humanGate
+    to: cancelled
+    guard:
+      kind: shell
+      options:
+        command: 'test "$__hitlAnswer" = "cancel"'
+terminalStates: [published, stop, cancelled]

package/config/workflows/idea-pipeline.yaml

@@ -145,6 +145,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Review the idea evaluation report (.spur/run/${vars.__runId}-idea-eval-report.md). Approve to create a feature, or reject to cancel? (Reject creates no feature.)"
+          decision:
+            mode: never
 
   - id: feature-create
     description: >
@@ -282,6 +284,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Feature check for ${vars.featureId}. Review the AC and confirm to proceed? Requirement coverage status: $(cat .spur/run/${vars.__runId}-idea-coverage.status 2>/dev/null || echo unknown) — reason: $(cat .spur/run/${vars.__runId}-idea-coverage.reason 2>/dev/null || echo n/a) (.spur/run/${vars.__runId}-idea-coverage.status[.reason]). (Failures route back to ac-generate for revision, capped at 3 retries.)"
+          decision:
+            mode: never
 
   - id: system-design
     description: >
@@ -358,6 +362,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Review the system design for feature ${vars.featureId}. Approve to proceed to decomposition? To reject, first record your feedback in .spur/run/${vars.__runId}-idea-design-review.md under the `## Operator feedback` heading, then answer no."
+          decision:
+            mode: never
 
   - id: decompose
     description: >
@@ -416,6 +422,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Batch-create for feature ${vars.featureId}. Review the task batch and confirm to proceed? (Failures route back to decompose for revision, capped at 3 retries.)"
+          decision:
+            mode: never
 
   - id: batch-create-run
     description: >

package/config/workflows/task-pipeline.yaml

@@ -464,6 +464,8 @@ states:
           - kind: hitl.confirm
             options:
                 prompt: "Approve task ${vars.wbs} to proceed to verification?"
+                decision:
+                    mode: never
 
     - id: verify
       description: >

package/config/workflows/wayfinder-resolution.yaml

@@ -163,6 +163,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Approve research resolution for task ${vars.wbs}?"
+          decision:
+            mode: never
 
   - id: record
     description: >

package/config/workflows/wrapup-pipeline.yaml

@@ -308,6 +308,8 @@ states:
       - kind: hitl.confirm
         options:
           prompt: "Branch cleanup for the validated task list (.spur/run/${vars.__runId}-wrapup-tasks.json). This is IRREVERSIBLE (merge or delete). Confirm to proceed?"
+          decision:
+            mode: never
 
   - id: done
     description: >

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@gobing-ai/spur",
-    "version": "0.3.89",
+    "version": "0.3.91",
     "description": "Spur CLI — local-first harness for mainstream coding agents: constraint checking, workflow orchestration, agent health, and history analytics. Bun-native; exposes the `spur` command.",
     "keywords": [
         "spur",
@@ -53,14 +53,14 @@
     },
     "devDependencies": {
         "@commander-js/extra-typings": "^14.0.0",
-        "@gobing-ai/ts-db": "^0.4.69",
-        "@gobing-ai/ts-ai-runner": "^0.4.69",
-        "@gobing-ai/ts-dual-workflow-engine": "^0.4.69",
-        "@gobing-ai/ts-infra": "^0.4.69",
-        "@gobing-ai/ts-llm-jsonl-importer": "^0.4.69",
-        "@gobing-ai/ts-rule-engine": "^0.4.69",
-        "@gobing-ai/ts-runtime": "^0.4.69",
-        "@gobing-ai/ts-utils": "^0.4.69",
+        "@gobing-ai/ts-db": "^0.5.2",
+        "@gobing-ai/ts-ai-runner": "^0.5.2",
+        "@gobing-ai/ts-dual-workflow-engine": "^0.5.2",
+        "@gobing-ai/ts-infra": "^0.5.2",
+        "@gobing-ai/ts-llm-jsonl-importer": "^0.5.2",
+        "@gobing-ai/ts-rule-engine": "^0.5.2",
+        "@gobing-ai/ts-runtime": "^0.5.2",
+        "@gobing-ai/ts-utils": "^0.5.2",
         "@types/bun": "1.3.14",
         "@types/figlet": "^1.7.0",
         "@types/node-notifier": "8.0.5",

package/plugins/sp/README.md

@@ -629,6 +629,7 @@ pipeline owns one lifecycle phase:
 | `idea-pipeline.yaml`        | Idea/planning → feature + tasks   | `/sp:dev-idea`, `/sp:dev-plan`    |
 | `wrapup-pipeline.yaml`      | Post-execution wrap-up            | `/sp:dev-wrap`, `/sp:dev-wrapall` |
 | `wayfinder-resolution.yaml` | Wayfinder ticket resolution loop  | `spur workflow run` (free-form)   |
+| `decision-routing-example.yaml` | DecisionMaker routing example (never / evidence / omitted modes, defer → operator pause) | `spur workflow run` (authoring sample) |
 
 ### Lifecycle operations
 

package/plugins/sp/agents/expert-spur.md

@@ -1,7 +1,7 @@
 ---
 name: expert-spur
-description: |
-  Use PROACTIVELY for "create tasks for this feature", "update all task statuses", "audit task traceability", "create a feature with acceptance criteria", "harden the rule catalog", "author a batch of workflows", "compose a workflow", "tune a rule", "apply accepted doctor proposals", "evaluate spur artifact health", "reflect over history findings", or "expert-spur". Multi-step corpus work across `spur task`, `feature`, `rule`, `workflow`, and agent specs: batch creation, status sweeps, section campaigns, traceability audits, rule hardening, workflow authoring/refactoring, composition and tuning (sp:spur-composer), and evaluation/reflection proposal passes (sp:spur-doctor). For one deterministic operation, run the CLI directly. Recurring evolution loops and multi-agent coordination belong to sp:super-planner — never this agent.
+description: |-
+  Use PROACTIVELY for "create tasks for this feature", "update all task statuses", "audit task traceability", or "expert-spur". Runs multi-step corpus campaigns across tasks, features, rules, workflows, and agent specs: creation, status/section updates, audits, and hardening. Uses sp:spur-composer for workflow composition, rule tuning, and accepted proposals; sp:spur-doctor for artifact evaluation and history reflection. Run single operations via CLI; recurring loops and multi-agent coordination belong to sp:super-planner.
 
   <example>
   Context: Batch task status update across a feature.