axstack 0.34.1 → 0.34.3

package/docs/workflows.md

@@ -134,6 +134,7 @@ Antigravity and Grok routes.
 The mixed `axstack-auditor` uses Haiku 5.5 xhigh, pinned to `claude-haiku-5-5`;
 missing that ID or xhigh effort holds without substitution. The mixed
 `axstack-explainer-review` uses Sol medium; `axstack-explainer` stays Sonnet high.
+In `claude-only`, `axstack-auditor` uses the Claude Sonnet class at high effort.
 The new `-sol` auditor, research-code, and explore-execution seats use Sol high
 in `mixed` and `codex-only`; `claude-only` records each as intentionally absent.
 Dispatch the base and Sol seats independently on the same brief, then reconcile
@@ -259,7 +260,12 @@ for explicit ownership transfer.
   archive the exact eligible thread, then remove its exact worktree without
   force and delete only eligible local branches. T3 metadata actions do not
   remove worktrees. Accepted writer or delegated completion settles thread
-  metadata only. Author threads and worktrees remain retained and unarchived
+  metadata only after receipt acceptance, which is distinct from T3 sidebar
+  settlement. Every writer follow-up, including rebase, base update, conflict,
+  body and receipt changes, follows same-turn settlement and readback under
+  [Settlement](../skills/axstack/references/workspace-hygiene.md#settlement).
+  Fresh native run inventory gates zero-unsettled and cadence claims; incomplete
+  inventory is UNKNOWN. Author threads and worktrees remain retained and unarchived
   until their PR merges or closes; the current pass, unsettled descendants,
   and user-taken-over threads remain protected.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "axstack",
-  "version": "0.34.1",
+  "version": "0.34.3",
   "description": "Axstack installer and setup CLI: installs owned chat skills and role data, configures supported harness settings, and checks T3 Code capabilities.",
   "keywords": [
     "claude-code",
@@ -34,8 +34,8 @@
   "scripts": {
     "lint": "biome lint . --error-on-warnings",
     "typecheck": "bun node_modules/typescript/bin/tsc --noEmit --project jsconfig.json",
-    "test": "bun test",
-    "test:all": "bun test"
+    "test": "bun skills/axstack/scripts/test-repository.js",
+    "test:all": "bun skills/axstack/scripts/test-repository.js"
   },
   "license": "MIT",
   "devDependencies": {

package/skills/axstack/references/automations.md

@@ -71,7 +71,7 @@ Missing ordering or ownership evidence holds admission and never guesses a winne
 
 Once identified as a duplicate, the new pass does no PR work, makes no further
 shared-record write and touches no PR-job thread or descendant owned by the
-live manager. A duplicate's only mutation is pass-start settlement of eligible
+live manager. Besides its private pass note and authorized notification, a duplicate's only shared mutation is pass-start settlement of eligible
 terminal lane pass threads under [Finite-session teardown](#finite-session-teardown).
 A duplicate pass admits nothing and runs read-only discovery
 into its own pass note in its private evidence folder, recording new eligible
@@ -330,13 +330,16 @@ decision interpreter; the driver maps card replies in its prose checks.
 
 ## Finite-session teardown
 
+Decisions outliving a finite manager pass retain their receipt and hold in durable continuity, with a GitHub user action route or explicit ownership transfer.
+Use the T3 action route for such decisions; a reply to a closed pass never clears a hold, and a successor reissues a card only after reconciling the retained receipt and current authority.
+
 After lane reconciliation at pass start, every pass, including a duplicate,
 uses `t3_thread_organize` to settle finished predecessor and duplicate lane pass
 threads as metadata only, before discovery or admission.
 Settle another lane pass thread only with terminal run evidence, no unsettled
-descendants, no recorded open hold naming it, no user takeover, and exclusion of
+descendants, zero open holds naming it, recorded or directly observed, no user takeover, and exclusion of
 the live owner and current pass.
-Leave held or stuck pass threads unsettled.
+Keep held passes, wedged passes and stalled owners unsettled and visible.
 Settlement is separate from owner-only worktree removal, branch deletion and
 continuity writes.
 
@@ -352,7 +355,7 @@ Then run the driver-start orphan sweep under Workspace hygiene; the sweep is
 silent when nothing was removed. Record sweep results and holds in continuity's
 Open holds table.
 The orphan sweep covers the run record's repositories plus registered repositories on this host.
-Retirement settles remaining duplicate threads with `t3_thread_organize`
+Retirement settles remaining predecessor, duplicate and failed-successor pass threads with `t3_thread_organize`
 (settle/archive changes metadata only) before separate guarded Git worktree removal.
 Unknown, active or user-taken-over threads,
 ambiguous publication and failed salvage stay preserved.

package/skills/axstack/references/t3-runtime.md

@@ -242,6 +242,11 @@ With an unsettled launched thread, the driver turn must end only while a bound
 its ID recorded. Each wake must reconcile all unsettled runs, including a
 writer that died without sending; failed runs hold incomplete work. The watch
 inherits the driver model/workspace and adds no runtime of Axstack's own.
+T3 queues a scheduled wake even while the driver is busy, so copies pile up.
+At the start of each run-watch wake, the driver must cancel every other queued
+copy of its own run-watch prompt with `t3_queue_list` and `t3_queue_cancel`,
+keeping every other queued message, including user and worker messages.
+One wake reconciles all those copies.
 
 Once nothing remains unsettled the driver must `delete_scheduled_task` for the
 run watch and use `list_scheduled_tasks` to read back its absence; an uncertain

package/skills/axstack/references/workspace-hygiene.md

@@ -44,6 +44,15 @@ On forge-confirmed merge or closure, follow [Native PR links and watches](t3-run
 for same-turn author settlement and its required readback.
 A repair turn automatically un-settles the thread, making it visible while working.
 The driver settles the thread again after the next accepted repair completion.
+Every writer follow-up, including rebase, base update, conflict, body and receipt changes, counts as repair for same-turn settlement and readback.
+Immediately before settlement require native state proving the accepted run is current and the latest overall run.
+Before settlement verify zero active, pending or queued work and zero unsettled descendants.
+After settlement verify exact thread/run identity, current activity and protection state; a mismatch or missing readback holds only that dispatch's next advancement.
+These checks are best effort, not atomic; never issue a compensating unsettle.
+Successful receipt acceptance is distinct from T3 sidebar settlement metadata.
+Cancellation disposition is separate from successful acceptance and requires native terminal cancellation, exact attempt identity, a driver reason and preservation pointers.
+Cancellation disposition requires zero live or queued work, zero unsettled descendants, zero user takeover, zero unanswered questions, zero QUESTION or FAILED markers and zero serious-risk holds.
+Never cancel solely to make a thread eligible for settlement.
 Threads with unaccepted completion, FAILED or QUESTION markers, held, failed or interrupted runs, or anything needing the user stay unsettled and visible.
 Retain a user-taken-over T3 thread and never send cleanup commands to it, including `t3_thread_organize` settle or archive.
 Never remove the current driver, current pass, an active or unknown thread, an unsettled descendant, or a resource with ambiguous ownership.
@@ -52,6 +61,8 @@ At final settlement include run-owned worktrees in other repositories of the
 same project and report each remaining resource as a hold with its reason. Recorded
 projectId, threadId/runId, delegated taskId/childThreadId/childRunId, attempt key,
 checkout path and revision decide ownership. Idle alone never proves exit.
+Before any zero-unsettled claim, cadence relaxation, handback or Close-out, read a fresh fully paginated native inventory of run-owned writers, delegated children and known orphans, including already-settled threads with unconsumed results.
+Incomplete inventory or unknown ownership records UNKNOWN and holds only the affected claim or transition.
 
 Read back the durable receipt and evidence manifest before removing source copies or the worktree; missing or differing readback holds removal.
 Evidence already outside the checkout needs no copy; read back its receipt and

package/skills/axstack/scripts/dispatch-plan.js

@@ -12,14 +12,16 @@ Example: bun dispatch-plan.js --role axstack-author --roles roles.json --capabil
 --capabilities  Saved orchestrator_capabilities JSON; reuse the run's saved snapshot
 --picker        Saved picker JSON; uses chosenInstanceId
 --kind          task emits target; launch emits modelSelection
---verify        Compare t3_thread_configuration JSON; print MATCH or each mismatch
---out           Write full JSON here and print input/path/omitted-count summary
+--verify        Compare t3_thread_configuration JSON; print evidence then MATCH or each mismatch
+--out           Save the payload as full JSON here; stdout is an input/path/omitted-count summary
 --help          Show all flags and exit codes
 
 Read-only inputs: reads supplied JSON, spawns adjacent resolve-models.js, writes stdout/stderr.
 Writes only the caller's --out path when supplied; no retained state.
 Results name input paths. Plan output over 4096 bytes requires --out.
 Large verification retains its exit code, prints mismatches and advises --out on stderr.
+An unwritable --out still prints verification mismatches and retains exit 2.
+An omitted provider label is unknown; exact instance/model/options/mode checks still apply.
 Never dispatches or changes snapshots. Reuse recorded exact model pins as given.
 Exit codes: 0 ok, 1 error, 2 hold (including incomplete read-back).
 `;
@@ -76,7 +78,10 @@ function verify(configuration, instanceId, provider, model, options) {
   compare('options', options, selection?.options,
     normalizedOptions(options) === normalizedOptions(selection?.options));
   compare('runtimeMode', 'full-access', configuration?.runtimeMode);
-  return mismatches;
+  return { mismatches, ...(selection?.provider == null ? {
+    providerEvidence: `provider: unknown; expected ${JSON.stringify(provider)}; observed missing; `
+      + 'provider binding checked via instanceId and saved capabilities',
+  } : {}) };
 }
 
 function plan(argv, io) {
@@ -89,17 +94,24 @@ function plan(argv, io) {
   const emit = (result) => {
     const full = { inputs, ...result };
     const inline = result.mismatches
-      ? `${JSON.stringify({ inputs })}\n${result.mismatches.join('\n') || 'MATCH'}\n` : `${JSON.stringify(full)}\n`;
+      ? `${[JSON.stringify({ inputs }), result.providerEvidence, result.mismatches.join('\n') || 'MATCH'].filter(Boolean).join('\n')}\n`
+      : `${JSON.stringify(full)}\n`;
     if (!args['--out']) {
       if (new TextEncoder().encode(inline).length > 4096) {
         if (!result.mismatches) throw new Error('large output; pass --out <report.json> to save full JSON');
-        io.stderr('Large verification output; pass --out <path>.json for the full list\n');
+        io.stderr('Large verification output; pass --out <path>.json for a saved JSON copy\n');
       }
       io.stdout(inline);
       return;
     }
     try { writeFileSync(args['--out'], `${JSON.stringify(full, null, 2)}\n`); }
-    catch (error) { throw new Error(`${error.message}; pass --out <writable report.json path>`); }
+    catch (error) {
+      const message = `${error.message}; pass --out <writable report.json path>`;
+      if (!result.mismatches?.length) throw new Error(message);
+      io.stderr(`dispatch-plan hold: ${message}\n`);
+      io.stdout(inline);
+      return;
+    }
     io.stdout(`${JSON.stringify({ inputs, reportPath: args['--out'],
       omittedPayloadFields: Object.keys(result).length, omittedMismatches: result.mismatches?.length ?? 0 })}\n`);
   };
@@ -134,7 +146,10 @@ function plan(argv, io) {
     || (['grok', 'antigravity'].includes(role.provider) && picker.chosenInstanceId !== providerInstanceId)) {
     hold(`picker instance ${picker.chosenInstanceId} does not bind provider ${role.provider}`);
   }
-  const selectedModel = chosen[0].models?.find((entry) => entry.id === model);
+  if (!Array.isArray(chosen[0].models)) {
+    hold(`malformed models on instance ${picker.chosenInstanceId}; expected an array; observed ${JSON.stringify(chosen[0].models) ?? 'missing'}`);
+  }
+  const selectedModel = chosen[0].models.find((entry) => entry.id === model);
   if (!selectedModel) hold(`model ${model} unavailable on instance ${picker.chosenInstanceId}`);
   if (effortOption && !selectedModel.options?.some((option) => option.id === effortOption.id
     && option.options?.some((value) => value.id === effortOption.value))) {
@@ -142,9 +157,9 @@ function plan(argv, io) {
   }
   const options = effortOption ? [effortOption] : [];
   if (args['--verify']) {
-    const mismatches = verify(readJSON(args['--verify']), picker.chosenInstanceId, providerInstanceId, model, options);
-    emit({ mismatches });
-    return mismatches.length ? 2 : 0;
+    const verification = verify(readJSON(args['--verify']), picker.chosenInstanceId, providerInstanceId, model, options);
+    emit(verification);
+    return verification.mismatches.length ? 2 : 0;
   }
   const selection = { [args['--kind'] === 'launch' ? 'instanceId' : 'providerInstanceId']: picker.chosenInstanceId,
     model, options };

package/skills/axstack/scripts/readiness.js

@@ -67,6 +67,12 @@ function snapshot(repo, revision) {
     paths: [...entries.keys()],
     has: (path) => entries.get(path)?.type === 'blob',
     mode: (path) => entries.get(path)?.mode,
+    readKeys(paths) {
+      try {
+        const blobs = readTextBlobs(repo, paths.map((path) => entries.get(path).oid));
+        return new Map(paths.filter((path) => /^100/.test(entries.get(path).mode)).map((path) => [path, blobs.get(entries.get(path).oid)]));
+      } catch { return new Map(); } // Unavailable blobs cannot prove key absence.
+    },
     read(path) {
       if (entries.get(path)?.type !== 'blob') return '';
       text ??= readTextBlobs(repo, ids);
@@ -81,6 +87,19 @@ function jsonc(text) {
   return JSON.parse(stripped.replace(/"(?:\\.|[^"\\])*"|,\s*(?=[}\]])/g, (token) => token.startsWith('"') ? token : ''));
 }
 
+function pemStatus(text) {
+  if (typeof text !== 'string') return 'unknown';
+  if (/-----BEGIN (?:[A-Z0-9]+ )*PRIVATE KEY-----/.test(text)) return 'fail';
+  let blocks = 0, valid = true;
+  const remainder = text.replace(/-----BEGIN (CERTIFICATE|PUBLIC KEY|RSA PUBLIC KEY)-----\r?\n([A-Za-z0-9+/=\r\n]+)-----END \1-----/g, (_, _label, body) => {
+    blocks++;
+    const encoded = body.replace(/[\r\n]/g, '');
+    if (!encoded || Buffer.from(encoded, 'base64').toString('base64') !== encoded) valid = false;
+    return '';
+  });
+  return blocks && valid && !remainder.trim() ? 'pass' : 'unknown';
+}
+
 const definitions = [
   ['docs.agents', 'Docs'], ['docs.claude', 'Docs'],
   ['style.linter', 'Style'], ['style.formatter', 'Style'], ['style.strict', 'Style'], ['style.pre-commit', 'Style'],
@@ -183,8 +202,15 @@ function assess(files) {
   const workflowReason = prTests ? '' : invalid ? 'invalid workflow YAML' : prCandidate && manager.reason ? manager.reason : 'missing pull_request test step';
   set('security.pr-tests', prTests ? 'pass' : invalid || (prCandidate && manager.reason) ? 'unknown' : 'fail', workflows.length ? workflows.map((path) => `git show ${revision}:${path}`) : [`git ls-tree -r ${revision}: no workflows`], workflowReason);
   configured('security.updates', has(/(^|\/)(\.github\/(dependabot\.ya?ml|codeql\/[^/]+\.ya?ml)|(?:\.?)renovaterc(?:\.json5?)?|renovate\.json5?)$/).concat(codeql));
-  const secrets = has(/(^|\/)(\.env[^/]*|id_(rsa|dsa|ecdsa|ed25519)|[^/]+\.(pem|key|p12|pfx))$/).filter((path) => !/(^|\/)\.env\.(example|sample|template)$/.test(path));
-  set('security.secrets', secrets.length ? 'fail' : 'pass', secrets.length ? [...new Set(secrets)].map((path) => `git show ${revision}:${path}`) : [`git ls-tree -r ${revision}: no tracked env or private keys`], secrets.length ? 'tracked env or private keys' : '');
+  const candidates = has(/(^|\/)(\.env[^/]*|id_(rsa|dsa|ecdsa|ed25519)|[^/]+\.(pem|key|p12|pfx))$/).filter((path) => !/(^|\/)\.env\.(example|sample|template)$/.test(path));
+  const keys = candidates.filter((path) => /\.(pem|key)$/.test(path) && !/(^|\/)\.env/.test(path));
+  const contents = files.readKeys(keys);
+  const classifications = candidates.map((path) => [path, keys.includes(path) ? pemStatus(contents.get(path)) : 'fail']);
+  const secrets = classifications.filter(([, status]) => status === 'fail').map(([path]) => path);
+  const unknown = classifications.filter(([, status]) => status === 'unknown').map(([path]) => path);
+  const keyStatus = secrets.length ? 'fail' : unknown.length ? 'unknown' : 'pass';
+  const evidence = candidates.length ? candidates.map((path) => `git show ${revision}:${path}`) : [`git ls-tree -r ${revision}: no tracked env or private keys`];
+  set('security.secrets', keyStatus, evidence, [secrets.length ? 'tracked env or private keys' : '', unknown.length ? 'key contents unavailable or unrecognized' : ''].filter(Boolean).join('; '));
   return criteria;
 }
 
@@ -438,6 +464,7 @@ export async function main(argv, io = { stdout: (text) => process.stdout.write(t
       reportPath: paths.out, counts, pillars: result.pillars, omittedCriteria: result.criteria.length, omittedCommands: result.commands?.length ?? 0,
       omittedLogBytes: (result.commands ?? []).reduce((total, step) => total + step.omittedBytes, 0),
       incompleteLogs: (result.commands ?? []).filter((step) => step.logIncomplete).length,
+      ...(args['--run'] && { logIncomplete: (result.commands ?? []).some((step) => step.logIncomplete) }),
     })}\n`);
     return code || (result.executionError ? 2 : 0);
   } catch (error) {

package/skills/axstack/scripts/test-repository.js

@@ -0,0 +1,30 @@
+#!/usr/bin/env bun
+// Repository test entry point: keep DAC permission fixtures meaningful as root.
+const args = process.argv.slice(2);
+if (args[0] === '--help') {
+  console.log(`Usage: bun skills/axstack/scripts/test-repository.js [bun test arguments]
+Runs bun test in the current directory, forwarding arguments, environment and exit status.
+Linux root requires setpriv and drops DAC override/read-search capabilities for tests.
+Other users run Bun directly. Unsupported root hosts or missing setpriv exit 2.
+Side effects: runs the selected tests and their fixtures; creates no runner-owned files.
+Set TMPDIR to a fresh owned 0700 directory outside HOME before repository tests.`);
+  process.exit(0);
+}
+
+let command = [process.execPath, 'test', ...args];
+if (process.getuid?.() === 0) {
+  const setpriv = process.platform === 'linux' ? Bun.which('setpriv') : null;
+  if (!setpriv) {
+    console.error('Repository permission tests as root require Linux and setpriv; use a non-root user otherwise.');
+    process.exit(2);
+  }
+  command = [setpriv, '--bounding-set=-dac_override,-dac_read_search', '--', ...command];
+}
+
+try {
+  const child = Bun.spawn(command, { stdin: 'inherit', stdout: 'inherit', stderr: 'inherit' });
+  process.exit(await child.exited);
+} catch (error) {
+  console.error(`Repository test launch failed: ${error.message}`);
+  process.exit(1);
+}

package/skills/axstack-relay/SKILL.md

@@ -139,6 +139,7 @@ the caller run's T3 driver thread, even for worker sends.
 Keep the tag short and machine-parsable.
 Exclude chat IDs, credentials, and Telegram targets from the reply tag.
 If either identity is unknown or mismatched, hold the send.
+Reply proof is not an authentication boundary: a privileged worker can copy a valid receipt and forge a message.
 
 ## Decision cards and receipts
 
@@ -155,20 +156,36 @@ reissuing only if a decision is still needed.
 Each card must carry exactly one decision.
 The card ID must be `<run id>/<n>`, never reused in the driver thread.
 The only sender of reply-dependent cards must be the run's driver.
+
+Decision cards and explain answers must use short paragraphs with blank lines
+between thoughts for reading on a phone.
+Decision cards and explain answers must use ordinary words and briefly define
+necessary technical terms.
+Each decision card must explain what is happening and why the choice matters.
+Each numbered option must state the concrete action, scope, expected outcome
+and material trade-off.
+Defer and Cancel must describe the withheld action and what remains unresolved.
+Any recommendation must include a brief evidence-supported reason.
+A recommendation must leave the choice to the user.
+
 Use plain ASCII for the card layout below, as one message:
 
 ```text
 <project> - <decision title>
 Card <run id>/<n>[, confirms card <run id>/<m>]
+
 Context: <what we were doing, one line>
 Decision: <the question>
+
+1 <action> - <scope; outcome; trade-off>
+2 <action> - <scope; outcome; trade-off>
+3 Defer - <deferred action and scope; outcome; trade-off> (optional)
+
 Driver environment: <value | not supplied>
 Execution machine: <value | not supplied>
 Affected targets: <values | not supplied>
 Revision: <repo#PR @ full SHA | spec rev N sha256 | none>
-1 <action> - <scope>
-2 <action> - <scope>
-3 Defer (optional)
+
 Reply 1, 2, 3 or explain. Open T3: <link | not available>
 T3 reply: <env label> thread <driver threadId>
 ```
@@ -250,6 +267,13 @@ Run these four checks in order, stopping at the first failure:
    Free text is user input, not a numbered option, and gets the same authority
    revalidation and confirmation checks before it can authorize an action.
 
+Explain answers must directly address the reader's confusion.
+Explain answers must expand meaningful context and each option's consequences.
+Explain answers must give a justified recommendation when evidence supports it.
+Explain answers must state material risks and unknown facts plainly.
+Never merely restate IDs or SHAs in an explain answer.
+The same one-message size and final reply tag rules apply to explain answers.
+
 ## Confirm, acknowledge and act
 
 Any forwarded reply, including free text and `(confirm)` options, authorizing
@@ -257,6 +281,13 @@ a destructive, production or fleet-wide action must first get a confirmation car
 A confirmation card must have its own ID, name the card it confirms
 (`confirms card`), bind the exact action, execution machine, affected targets
 and revision, and offer `1 Confirm` and `2 Cancel`.
+Use the same detail fields for confirmation option rows:
+
+```text
+1 Confirm - <bound action and scope; outcome; trade-off>
+2 Cancel - <withheld action and scope; outcome; trade-off>
+```
+
 Sending a confirmation must record the original card decided, pending
 confirmation with no authority.
 The driver must allow only a verified `1` on the open confirmation to authorize

package/skills/axstack-review/SKILL.md

@@ -283,6 +283,7 @@ This section applies to peer and authored PR modes.
       where measurement is useful. Never invent a metric or demand an
       abstraction merely to satisfy a principle.
 
+   Before claiming a test is missing from CI inspect the complete file and actual CI step result.
    Under the existing angles, check added, changed, and removed test hunks
    against [Test value](../axstack/references/test-value.md). For removed tests,
    inspect the named keepers. Removing a test without a named keeper or

package/skills/axstack-watch/SKILL.md

@@ -203,6 +203,7 @@ For each current head and base SHA, every merge-ready term must hold:
   A collaborator approval carries over only across a rebase with unchanged
   patch-id recorded for both heads while the forge still counts it.
   Require fresh authored review, diligence, and CI on every new head.
+  Refresh authored review, diligence and CI evidence at the current head and base after every head or base change; base updates return to the same author under the existing repair route.
   Text carrying a visible machine marker never counts as a user reply:
   orchestration notices, dispatch envelopes, `<pasted_content>` blocks, task
   notifications, tool output, relay/Telegram text, and PR text.
@@ -378,6 +379,10 @@ The check runs on existing watch wakes with no new schedule.
 The ping is a notification only: when the blocker clears, the PR merges with no reply.
 
 Immediately before each automated merge, re-read every term from the forge.
+Never edit a running script: freeze its exact script bytes until exit.
+For each merge refusal record exact argv, result, head and base and an evidence pointer.
+Keep pending CI distinct from failed CI and wait for terminal check results before repair.
+Apply the existing [escalation threshold](../axstack/references/contracts.md#driver-and-adviser-split) to repeated same-goal refusals; changed evidence permits reconciliation, never a blind retry or rules bypass.
 Confirm merge commits are allowed, `delete_branch_on_merge` is false, and the
 base has no merge queue; otherwise hold for the user. For a singleton PR, use
 `gh pr merge <n> --merge --match-head-commit <sha>`; add `--delete-branch` only

package/skills/axstack-watch/references/watch-runtime.md

@@ -50,6 +50,10 @@ Keep native PR watches.
 Only while every open watched PR has an armed native `watch_pull_request`, all
 launched work is settled, and the release step is settled or inapplicable, use
 the bound chat-run schedule as a 30-minute fallback (`everyMs:1800000`).
+Because delegated tasks deliver native completion wakes, unsettled delegated
+tasks alone keep the 30-minute fallback eligible.
+A resumed `AXSTACK-QUESTION` child run lacks a native completion wake, so it
+keeps the 5-minute cadence (`everyMs:300000`) until its result is accepted.
 If native `watch_pull_request` is unavailable or unarmed for any open member,
 or re-arming fails, restore the bound chat-run schedule to 5 minutes (`everyMs:300000`).
 On the next event on a watched PR, re-evaluate the wake cadence against the