merge upstream master into COD-358

docs/superpowers/plans/2026-08-23-workflow-run-watcher-clock-independent-test.md

@@ -1,69 +0,0 @@
-# Workflow Run Watcher Clock-Independent Test Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Prevent `workflow-run-watcher.test.ts` from expiring as wall-clock time advances.
-
-**Architecture:** Keep production code unchanged. Anchor every synthetic workflow timestamp to one captured current
-time while preserving the fixture's existing event offsets, then verify the recent-summary assertion against the
-real watcher.
-
-**Tech Stack:** TypeScript, Vitest, Node.js
-
----
-
-### Task 1: Make the completed-run fixture relative to test time
-
-**Files:**
-
-- Modify: `test/workflow-run-watcher.test.ts:17-100`
-
-- [x] **Step 1: Confirm the fixed fixture fails after its recent window expires**
-
-Run:
-
-```bash
-npm test -- test/workflow-run-watcher.test.ts
-```
-
-Expected: FAIL in `getRecentRunSummaries omits agents[] (lightweight snapshot)` because `summaries` is empty.
-
-- [x] **Step 2: Anchor the fixture to the current test time**
-
-At the start of `sampleRunJson()`, capture a start time fifteen minutes before the call:
-
-```typescript
-const startTime = Date.now() - 15 * 60_000;
-```
-
-Replace the fixed ISO timestamp with `new Date(startTime).toISOString()`, replace the fixed `startTime` property
-with the variable, and express each workflow agent's `startedAt`, `queuedAt`, and `lastProgressAt` as its existing
-millisecond offset from `startTime`.
-
-- [x] **Step 3: Verify the focused test passes**
-
-Run:
-
-```bash
-npm test -- test/workflow-run-watcher.test.ts
-```
-
-Expected: 19 tests pass with no failures.
-
-- [x] **Step 4: Verify formatting and the CI gate**
-
-Run:
-
-```bash
-npx prettier --check test/workflow-run-watcher.test.ts
-npm run test:ci
-```
-
-Expected: formatting passes and the CI unit/integration suite has no failures.
-
-- [x] **Step 5: Commit the correction**
-
-```bash
-git add test/workflow-run-watcher.test.ts docs/superpowers/plans/2026-08-23-workflow-run-watcher-clock-independent-test.md
-git commit -m "test(workflows): keep recent-run fixture clock-independent"
-```

docs/superpowers/specs/2026-08-23-workflow-run-watcher-clock-independent-test-design.md

@@ -1,24 +0,0 @@
-# Workflow Run Watcher Clock-Independent Test Design
-
-## Problem
-
-`getRecentRunSummaries omits agents[] (lightweight snapshot)` uses a workflow fixture whose activity timestamps
-are fixed in June 2026. The test requests a 100,000-minute recent window, so it began failing once wall-clock time
-moved beyond that window even though the implementation had not changed.
-
-## Design
-
-Keep production code unchanged. Build the synthetic run from one captured `Date.now()` value and express its
-timestamp, start time, queue times, progress times, and completion times as offsets from that value. Preserve the
-existing ordering and duration relationships between workflow events, so the fixture remains representative while
-its completed run always falls within the test's recent window.
-
-Do not widen the window or replace `getRecentRunSummaries()` with an unfiltered API: either choice would weaken or
-eventually reintroduce the regression. Do not install fake timers, because the watcher uses asynchronous discovery
-and timer behavior that this assertion does not need to control.
-
-## Verification
-
-- Confirm the existing fixed fixture fails because the run falls outside the recent window.
-- Run `test/workflow-run-watcher.test.ts` after converting the fixture to relative timestamps.
-- Run formatting and the CI unit/integration gate before pushing the updated PR.

test/workflow-run-watcher.test.ts

@@ -17,13 +17,22 @@ const PROJECT_HASH = '-home-arkon-default-claudeman';
 const SESSION_UUID = '388113c8-cd01-4e80-93a8-3be66ab1519b';
 const RUN_ID = 'wf_test1234-abc';
 
+/**
+ * The fixture's epochs are anchored to "now", never pinned, because the recency
+ * assertions below compare them against `Date.now()`. A frozen epoch plus a fixed
+ * window is a time bomb: the original fixture's newest activity sat at
+ * 2026-06-14T20:06:40Z, and `getRecentRunSummaries(100000)` — that argument is
+ * MINUTES, i.e. 69.4 days — stopped matching it on 2026-08-23T06:46:40Z, turning
+ * CI red on a suite nobody had touched. Offsets from the anchor are preserved
+ * verbatim, so every parsed duration and ordering assertion is unchanged.
+ */
+const RUN_ANCHOR = Date.now() - 601_000;
+
 /** A run JSON shaped like a real (killed) run: all three agent states + the bloat fields. */
 function sampleRunJson() {
-  const startTime = Date.now() - 15 * 60_000;
-
   return {
     runId: RUN_ID,
-    timestamp: new Date(startTime).toISOString(),
+    timestamp: new Date(RUN_ANCHOR).toISOString(),
     taskId: 'task_abc',
     // --- bloat fields that MUST be stripped ---
     script: 'export const meta = {};\n'.repeat(5000), // ~110KB
@@ -37,7 +46,7 @@ function sampleRunJson() {
     workflowName: 'review-open-prs',
     status: 'killed',
     error: 'user stopped the task',
-    startTime,
+    startTime: RUN_ANCHOR,
     defaultModel: 'claude-opus-4-8[1m]',
     totalTokens: 109703,
     totalToolCalls: 44,
@@ -57,13 +66,13 @@ function sampleRunJson() {
         agentId: 'a6c0e282c3f5ac0bf',
         model: 'claude-opus-4-8[1m]',
         state: 'done',
-        startedAt: startTime + 1002,
-        queuedAt: startTime + 962,
+        startedAt: RUN_ANCHOR + 1002,
+        queuedAt: RUN_ANCHOR + 962,
         attempt: 1,
         lastToolName: 'StructuredOutput',
         lastToolSummary: 'Does the profile setting make the allowlist dead config',
         promptPreview: 'You are reviewing a pull request...',
-        lastProgressAt: startTime + 525143,
+        lastProgressAt: RUN_ANCHOR + 525_143,
         tokens: 104703,
         toolCalls: 41,
         durationMs: 524140,
@@ -78,12 +87,12 @@ function sampleRunJson() {
         agentId: 'a1234567890abcdef',
         model: 'claude-opus-4-8[1m]',
         state: 'progress',
-        startedAt: startTime + 11000,
-        queuedAt: startTime + 970,
+        startedAt: RUN_ANCHOR + 11_000,
+        queuedAt: RUN_ANCHOR + 970,
         attempt: 1,
         lastToolName: 'Read',
         promptPreview: 'Review PR 127...',
-        lastProgressAt: startTime + 601000,
+        lastProgressAt: RUN_ANCHOR + 601_000,
         tokens: 5000,
         toolCalls: 3,
       },
@@ -95,9 +104,9 @@ function sampleRunJson() {
         phaseTitle: 'Verify',
         model: 'claude-opus-4-8[1m]',
         state: 'start',
-        queuedAt: startTime + 980,
+        queuedAt: RUN_ANCHOR + 980,
         promptPreview: 'Verify finding x...',
-        lastProgressAt: startTime + 980,
+        lastProgressAt: RUN_ANCHOR + 980,
       },
     ],
   };