S00-025: Verify storage recovery launcher after argument repair

Changes: Record reproduction of the S00-023 argument conflict and verification of the active reviewed S00-028 invocation; replace blocked S00-025 handoff with bounded QA instructions. Preserve all storage source artifacts.
Validation: Conflicting invocation returns expected exit 2; current argv help and version checks return 0; tasks.py validate passes. No model, storage or browser correctness test performed.

Task: S00-025
Agent: auto-storage-8b47cee148df
Role: storage
Action: claim
Status: review

Task title: Устранить прерывание: Устранить прерывание: Выбрать IndexedDB wrapper и migration policy

state/commit_messages/auto-storage-8b47cee148df.txt

@@ -0,0 +1,4 @@
+Verify storage recovery launcher after argument repair
+
+Changes: Record reproduction of the S00-023 argument conflict and verification of the active reviewed S00-028 invocation; replace blocked S00-025 handoff with bounded QA instructions. Preserve all storage source artifacts.
+Validation: Conflicting invocation returns expected exit 2; current argv help and version checks return 0; tasks.py validate passes. No model, storage or browser correctness test performed.

state/evidence/S00-025-auto-storage-8b47cee148df.md

@@ -0,0 +1,56 @@
+# S00-025 executor evidence
+
+Agent: auto-storage-8b47cee148df; role: storage; 2026-10-15.
+
+Input: assigned card/packet, required startup/workflow reads, TESTING, ADR registry,
+blockers S00-023/S00-025, prior S00-025 and parent handoffs, accepted ADR 0001 and
+S00-002/S00-028 executor/review evidence and handoffs.
+Output: verified removal of the specific launcher interruption; no storage implementation.
+Acceptance: historical argument error reproduced, current invocation matches reviewed repair,
+original storage task remains unaccepted and ready for its separate fresh recovery cycle.
+
+## Findings and commands
+
+- resume --agent auto-storage-8b47cee148df, show S00-025 and packet S00-025
+  --agent auto-storage-8b47cee148df via python3 scripts/tasks.py: exit 0 each;
+  assigned executor, dependencies S00-002/S00-028 done, no repeat claim.
+- Read original state/runs/auto-storage-87486895c792/run.json and output.log:
+  exit 0; historical S00-023 process exited 2 before execution with the mutually
+  exclusive --approve-for-me/--sandbox diagnostic.
+- Python subprocess.run(['codex','exec','--approve-for-me','--sandbox',
+  'workspace-write','-'], input='', capture_output=True, text=True, timeout=20):
+  child exit 2 and matching "cannot be used with" diagnostic; assertions/wrapper exit 0.
+  Parser rejects before model execution. Read-only PATH alias warning also present.
+- Python JSON assertions on runner.local.json: argv exactly
+  ['codex','-c','sandbox_workspace_write.network_access=true','exec',
+  '--approve-for-me','--ephemeral','-']; commands_by_role empty. No config edits.
+  subprocess.run(argv+['--help'], input='', capture_output=True, text=True,
+  timeout=20): child exit 0; stdout contains workspace-write; wrapper exit 0.
+- codex --version through timeout20 subprocess: exit 0, codex-cli 0.159.0.
+- An initial negative check appended --sandbox workspace-write AND --help:
+  child exit 0, wrapper assertion exit 1. Help short-circuits conflict validation;
+  this is not successful negative evidence. Replaced with the no-help rejection
+  above, which reproduces the historical error without launching a model.
+- Initial guessed docs/decisions/0001-web-stack.md read returned exit 1;
+  correct existing docs/decisions/0001-stack.md located through dependency evidence
+  and read. No required artifact missing.
+- python3 scripts/tasks.py validate: exit 0, plan/state validation passed.
+- Read-only git status --short initially: only coordinator claim state/tasks.json.
+  Final git diff --check: exit 0. git status --short --untracked-files=all:
+  exit 0; only state/tasks.json coordinator claim, this evidence, S00-025 handoff
+  and agent commit message changed. No source changes outside scope.
+
+OpenAI Docs skill used. Official CLI reference searched and opened on 2026-10-15:
+https://learn.chatgpt.com/docs/cli/reference ; documents approve-for-me as automatic
+review with workspace-write. Exact local results above establish this host's parser behavior.
+No extra model, inference/authentication test or nested sandbox launched. The independent
+filesystem verification is existing accepted S00-028 evidence, not rerun or claimed here.
+The present fresh assignment itself received complete task inputs after that repair.
+
+## Scope and limits
+
+Only evidence, handoff and mandatory commit message authored; no source/config/ADR,
+partial storage probe, shared contract, immutable spec or plan definition changed.
+No wrapper/migration/quota/browser correctness accepted. S00-005 and S00-023 are not
+completed by this session. Their remaining work belongs to fresh assigned executors
+following independent review of S00-025. Coordinator alone commits and launches sessions.

state/handoffs/S00-025.md

@@ -1,13 +1,24 @@
-# S00-025 blocked handoff
+# S00-025 handoff for independent QA
 
-Agent: auto-storage-ae5930b6fbdf; role: storage.
+Executor: auto-storage-8b47cee148df; role: storage; 2026-10-15.
 
-Input: required workflow, storage role, S00-025 card, blocker reports S00-023/S00-025, parent handoffs and accepted S00-002 ADR/evidence.
+Specific cause: S00-023 launcher rejected --approve-for-me together with --sandbox,
+exit 2 before task execution. Reproduced the same parser error without --help and
+without launching a model. S00-028 already repaired and independently reviewed the
+coordinator invocation; active argv matches it exactly and --help exits 0 on Codex
+0.159.0. Current fresh assignment received complete inputs. No source fix needed here.
 
-Reached result: diagnosed the exact previous interruption through the runner's actual diagnostic. S00-023 harness exited before task execution: codex exec rejected mutually exclusive --approve-for-me and --sandbox flags. Repairing coordinator argv is outside S00-025 storage-only scope.
+Evidence: state/evidence/S00-025-auto-storage-8b47cee148df.md records exact commands,
+failed exploratory help assertion, successful reproduction and limitations.
+Plan/state validate passed. Only workflow evidence/handoff/message changed.
 
-Reproduction/evidence: state/runs/auto-storage-87486895c792/run.json records returncode=2 and identical start/finish timestamp. output.log reports: error: the argument '--approve-for-me' cannot be used with '--sandbox <SANDBOX_MODE>'. No nested model launched to reproduce.
+QA next: inspect executor diff/message, repeat the rejecting conflicting invocation
+with empty stdin/timeout20 and the active argv + --help check. Help alone cannot
+exercise the negative conflict case. Inspect accepted S00-028 sandbox evidence;
+no new model or original storage work is needed for this launcher remediation.
 
-The existing storage package and partial probes were read only. No ADR exists yet at docs/decisions/0004-storage-design.md; no storage selection, rollback correctness, quota handling or browser readiness is accepted. The partial rollback probe and missing quota script belong to subsequent fresh storage work, not this launcher repair.
-
-Next input: coordinator must supply a compatible codex exec argv preserving project write sandbox, removing the conflicting approval option. A fresh bounded ci-cd remediation must validate argument compatibility without launching product tasks, then receive independent review before any original task resumes. No runner configuration or source edits performed in this session.
+After independent acceptance and coordinator commit, S00-023 returns to a fresh
+executor to inspect preserved partial storage artifacts for its interruption recovery.
+S00-005 still owns the IndexedDB wrapper/migration decision and required experiments.
+No ADR 0004, rollback, quota or browser storage readiness is accepted by S00-025.
+Do not treat removal of this parser error as completion of either parent task.

state/tasks.json

@@ -1,6 +1,6 @@
 {
   "schema_version": 1,
-  "revision": 106,
+  "revision": 108,
   "tasks": {
     "S00-001": {
       "status": "done",
@@ -1359,18 +1359,19 @@
       "completed_at": "2026-10-15T10:04:38+00:00"
     },
     "S00-025": {
-      "status": "todo",
-      "owner": null,
+      "status": "review",
+      "owner": "auto-storage-8b47cee148df",
       "external_wait": false,
       "reason": null,
-      "started_at": "2026-10-15T09:46:27+00:00",
-      "updated_at": "2026-10-15T10:55:14+00:00",
+      "started_at": "2026-10-15T10:57:38+00:00",
+      "updated_at": "2026-10-15T10:59:32+00:00",
       "blocked_by": [],
       "resume_status": "in_progress",
       "blocker_report": "state/blockers/S00-028.md",
       "handoff": "state/handoffs/S00-025.md",
       "previous_owner": "auto-storage-ae5930b6fbdf",
-      "resumed_after_blocker": true
+      "resumed_after_blocker": true,
+      "evidence": "state/evidence/S00-025-auto-storage-8b47cee148df.md"
     },
     "S00-026": {
       "status": "done",
@@ -2805,6 +2806,29 @@
       "to": "done",
       "reason": null,
       "evidence": "state/evidence/S00-028-review.md"
+    },
+    {
+      "revision": 107,
+      "at": "2026-10-15T10:57:38+00:00",
+      "action": "claim",
+      "task": "S00-025",
+      "agent": "auto-storage-8b47cee148df",
+      "role": "storage",
+      "from": "todo",
+      "to": "in_progress",
+      "reason": null
+    },
+    {
+      "revision": 108,
+      "at": "2026-10-15T10:59:32+00:00",
+      "action": "submit",
+      "task": "S00-025",
+      "agent": "auto-storage-8b47cee148df",
+      "role": null,
+      "from": "in_progress",
+      "to": "review",
+      "reason": null,
+      "evidence": "state/evidence/S00-025-auto-storage-8b47cee148df.md"
     }
   ],
   "retired_agents": [