CC: Cover Codex retention in CTO validator

2026-06-04 14:59:18 -04:00
parent 68f071e1e6
commit a32a996e69
1 changed files with 160 additions and 0 deletions
@@ -111,6 +111,14 @@ REQUIRED_FILES = [
    ".sot/03-PROTOCOLS/CTO-CASE-SPARK-ENDPOINT-CONFIG-PRD.md",
    ".sot/03-PROTOCOLS/CTO-CASE-SPARK-ENDPOINT-CONFIG-ISSUES.md",
    ".sot/03-PROTOCOLS/CTO-CASE-AGENT-PROTOCOL-BLOCKER.md",
+    ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-DRY-RUN-PACKET.md",
+    ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-POLICY-PACKET.md",
+    ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-ARCHIVE-EXECUTOR-PACKET.md",
+    "tools/report_codex_retention_pressure.py",
+    "tools/plan_codex_retention_policy.py",
+    "tools/archive_codex_inactive_threads.py",
+    "tools/probe_codex_native_retention.py",
+    "tools/codex_ephemeral_exec.py",
 ]

 REQUIRED_BRIEF_PHRASES = [
@@ -1656,6 +1664,105 @@ REQUIRED_CORE_ROUTE_ADMISSION_GUARD_CLOSEOUT_PHRASES = [
    "candidate-only until the guard passes",
 ]

+REQUIRED_CODEX_RETENTION_ISSUE_IDS = [
+    "CTO-WORK-093",
+    "CTO-WORK-094",
+    "CTO-WORK-095",
+    "CTO-WORK-096",
+]
+
+REQUIRED_CODEX_RETENTION_DRY_RUN_PHRASES = [
+    "Local planning SOT only. Not a Core Protocol. Not active Core authority.",
+    "Codex retention pressure is measurable without reading raw transcripts or mutating `~/.codex`.",
+    "metadata-only JSON",
+    "It does not read transcript bodies, update SQLite, delete files, vacuum databases, archive threads, mutate Core, start Runtime, read secrets, change Codex config, or claim product readiness.",
+    "Prevention default: use `codex exec --ephemeral` for disposable non-interactive worker runs.",
+    "Blocked without explicit operator approval:",
+    "direct `threads.archived` updates",
+    "session JSONL deletion",
+    "logs table deletion",
+    "SQLite vacuum or checkpoint",
+    "raw transcript import into Core",
+    "broad cleanup of `~/.codex`",
+]
+
+REQUIRED_CODEX_RETENTION_POLICY_PHRASES = [
+    "Local planning SOT only. Not a Core Protocol. Not active Core authority.",
+    "Codex retention cleanup now has a backup-first, approval-gated policy plan.",
+    "`python3 tools/plan_codex_retention_policy.py` emits metadata-only JSON.",
+    "`python3 tools/probe_codex_native_retention.py` checks installed Codex CLI help, feature flags, and local version cache.",
+    "native cleanup/archive/retention command advertised by installed CLI: false",
+    "prevention flag advertised: `codex exec --ephemeral`",
+    "Phase 1: backup `state_5.sqlite`, `logs_2.sqlite`, WAL, and SHM files.",
+    "Phase 2: archive-only candidate threads by DB flag only after explicit approval.",
+    "Phase 3: delete archived session JSONL only after separate destructive approval.",
+    "`python3 tools/codex_ephemeral_exec.py` builds disposable worker commands as `codex exec --ephemeral`.",
+    "The helper is prevention only. It does not archive threads, delete JSONL, truncate logs, checkpoint, vacuum, read transcript bodies, or mutate Core.",
+    "Blocked without explicit operator approval:",
+    "updating `threads.archived`",
+    "deleting session JSONL",
+    "SQLite checkpoint or vacuum",
+    "raw transcript read/import",
+]
+
+REQUIRED_CODEX_RETENTION_ARCHIVE_PHRASES = [
+    "Local planning SOT only. Not a Core Protocol. Not active Core authority.",
+    "Codex retention cleanup now has a guarded archive-only executor.",
+    "Default mode is dry-run.",
+    "Mutation requires an exact approval token.",
+    "It does not delete session JSONL, truncate logs, checkpoint, vacuum, read transcript bodies, or import transcripts into Core.",
+    "python3 tools/archive_codex_inactive_threads.py --check",
+    "python3 tools/probe_codex_native_retention.py --check",
+    "I approve CTO-WORK-095 archive-only Codex threads older than 7 days.",
+    "candidate selection reads only `id`, `rollout_path`, `updated_at`, `archived`, and file size",
+    "mutation is limited to `threads.archived=1` and `archived_at`",
+    "Still blocked without separate approval:",
+    "delete archived session JSONL",
+    "run SQLite checkpoint or vacuum",
+    "read raw transcript bodies",
+]
+
+REQUIRED_CODEX_RETENTION_TOOL_PHRASES = {
+    "tools/report_codex_retention_pressure.py": [
+        "metadata_only",
+        "raw_transcript_bodies_read",
+        "mutation_performed",
+        "blocked_without_approval",
+    ],
+    "tools/plan_codex_retention_policy.py": [
+        "metadata_only",
+        "raw_transcript_bodies_read",
+        "raw_thread_text_fields_read",
+        "mutation_performed",
+        "approval_boundaries",
+        "false_effects",
+    ],
+    "tools/archive_codex_inactive_threads.py": [
+        "approval_token_invalid",
+        "backup_codex_state",
+        "raw_transcript_bodies_read",
+        "session_jsonl_deleted",
+        "sqlite_checkpoint_or_vacuum",
+        "false_effects",
+    ],
+    "tools/probe_codex_native_retention.py": [
+        "metadata_only",
+        "mutation_performed",
+        "native_retention_cleanup_command_advertised",
+        "ephemeral_prevention_advertised",
+        "false_effects",
+    ],
+    "tools/codex_ephemeral_exec.py": [
+        "codex",
+        "exec",
+        "--ephemeral",
+        "--print-command",
+        "dangerously-bypass-approvals-and-sandbox",
+        "session_persistence_requested",
+        "false_effects",
+    ],
+}
+

 def workboard_status(text: str, issue_id: str) -> str | None:
    pattern = rf"- id: {re.escape(issue_id)}\n(?:    .+\n)*?    status: ([^\n]+)"
@@ -2758,6 +2865,45 @@ def main() -> int:
            if phrase not in text:
                errors.append(f"missing_provider_decision_record_phrase:{phrase}")

+    codex_retention_dry_run = ROOT / ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-DRY-RUN-PACKET.md"
+    if codex_retention_dry_run.is_file():
+        text = codex_retention_dry_run.read_text(encoding="utf-8")
+        for phrase in REQUIRED_CODEX_RETENTION_DRY_RUN_PHRASES:
+            checked.append(f"codex_retention_dry_run_phrase:{phrase}")
+            if phrase not in text:
+                errors.append(f"missing_codex_retention_dry_run_phrase:{phrase}")
+        if "source: CTO-WORK-093" not in text:
+            errors.append("codex_retention_dry_run_missing_source")
+
+    codex_retention_policy = ROOT / ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-POLICY-PACKET.md"
+    if codex_retention_policy.is_file():
+        text = codex_retention_policy.read_text(encoding="utf-8")
+        for phrase in REQUIRED_CODEX_RETENTION_POLICY_PHRASES:
+            checked.append(f"codex_retention_policy_phrase:{phrase}")
+            if phrase not in text:
+                errors.append(f"missing_codex_retention_policy_phrase:{phrase}")
+        if "source: CTO-WORK-094" not in text:
+            errors.append("codex_retention_policy_missing_source")
+
+    codex_retention_archive = ROOT / ".sot/03-PROTOCOLS/CTO-CODEX-RETENTION-ARCHIVE-EXECUTOR-PACKET.md"
+    if codex_retention_archive.is_file():
+        text = codex_retention_archive.read_text(encoding="utf-8")
+        for phrase in REQUIRED_CODEX_RETENTION_ARCHIVE_PHRASES:
+            checked.append(f"codex_retention_archive_phrase:{phrase}")
+            if phrase not in text:
+                errors.append(f"missing_codex_retention_archive_phrase:{phrase}")
+        if "source: CTO-WORK-095" not in text:
+            errors.append("codex_retention_archive_missing_source")
+
+    for rel, phrases in REQUIRED_CODEX_RETENTION_TOOL_PHRASES.items():
+        path = ROOT / rel
+        if path.is_file():
+            text = path.read_text(encoding="utf-8")
+            for phrase in phrases:
+                checked.append(f"codex_retention_tool_phrase:{rel}:{phrase}")
+                if phrase not in text:
+                    errors.append(f"missing_codex_retention_tool_phrase:{rel}:{phrase}")
+
    board = ROOT / "WORKBOARD.yaml"
    if board.is_file():
        text = board.read_text(encoding="utf-8")
@@ -2825,6 +2971,10 @@ def main() -> int:
            checked.append(f"workboard_id:{issue_id}")
            if issue_id not in text:
                errors.append(f"missing_workboard_id:{issue_id}")
+        for issue_id in REQUIRED_CODEX_RETENTION_ISSUE_IDS:
+            checked.append(f"workboard_id:{issue_id}")
+            if issue_id not in text:
+                errors.append(f"missing_workboard_id:{issue_id}")
        expected_statuses = {
            "CTO-WORK-001": "validated",
            "CTO-WORK-002": "validated",
@@ -2915,6 +3065,10 @@ def main() -> int:
            "CTO-WORK-090": "validated",
            "CTO-WORK-091": "validated",
            "CTO-WORK-092": "validated",
+            "CTO-WORK-093": "validated",
+            "CTO-WORK-094": "validated",
+            "CTO-WORK-095": "validated",
+            "CTO-WORK-096": "validated",
        }
        for issue_id, expected in expected_statuses.items():
            checked.append(f"workboard_status:{issue_id}:{expected}")
@@ -3009,6 +3163,12 @@ def main() -> int:
            errors.append("workboard_missing_core_route_admission_guard_source")
        if "CTO-CORE-ROUTE-ADMISSION-GUARD-CLOSEOUT.md" not in text:
            errors.append("workboard_missing_core_route_admission_guard_closeout_source")
+        if "CTO-CODEX-RETENTION-DRY-RUN-PACKET.md" not in text:
+            errors.append("workboard_missing_codex_retention_dry_run_source")
+        if "CTO-CODEX-RETENTION-POLICY-PACKET.md" not in text:
+            errors.append("workboard_missing_codex_retention_policy_source")
+        if "CTO-CODEX-RETENTION-ARCHIVE-EXECUTOR-PACKET.md" not in text:
+            errors.append("workboard_missing_codex_retention_archive_source")

    payload = {
        "ok": not errors,