arkaos 5.17.0 → 5.17.2

package/README.md

@@ -4,7 +4,7 @@
 
 **The Operating System for AI Agent Teams.**
 
-89 agents. 17 departments. 340 skills. Enterprise frameworks. Multi-runtime. One install.
+89 agents. 17 departments. 341 skills. Enterprise frameworks. Multi-runtime. One install.
 
 ```bash
 npx arkaos install
@@ -97,7 +97,7 @@ npx arkaos doctor    # Health check
 
 ### Skill packs, à la carte
 
-The default install ships a curated core so your context window stays lean. Everything else lives in the ArkaOS plugin marketplace: 16 department packs with 266 skills, generated straight from the same sources the core uses. Inside Claude Code:
+The default install ships a curated core so your context window stays lean. Everything else lives in the ArkaOS plugin marketplace: 16 department packs with 267 skills, generated straight from the same sources the core uses. Inside Claude Code:
 
 ```
 /plugin marketplace add andreagroferreira/arka-os

package/THE-ARKAOS-GUIDE.md

@@ -1,6 +1,6 @@
 # The ArkaOS Guide
 
-> v5.16.1 — 89 agents, 17 departments, 340 skills, 306 commands, 20 ADRs.
+> v5.17.1 — 89 agents, 17 departments, 341 skills, 307 commands, 20 ADRs.
 > One file, everything you need to start. Generated by `scripts/guide_gen.py` — never hand-edited.
 
 ## What it is
@@ -40,7 +40,7 @@ answers as a generic assistant.
 | `/arka` | 22 |
 | `/brand` | 14 |
 | `/community` | 15 |
-| `/content` | 20 |
+| `/content` | 21 |
 | `/dev` | 42 |
 | `/do` | 1 |
 | `/ecom` | 18 |

package/VERSION

@@ -1 +1 @@
-5.17.0
+5.17.2

package/arka/SKILL.md

@@ -63,7 +63,7 @@ complementary, not alternatives.
 # ArkaOS — Main Orchestrator
 
 > **The Operating System for AI Agent Teams**
-> 89 agents. 17 departments. 340 skills. Multi-runtime. Dashboard. Knowledge RAG.
+> 89 agents. 17 departments. 341 skills. Multi-runtime. Dashboard. Knowledge RAG.
 
 ## ⛔ Evidence flow — 4 gates (NON-NEGOTIABLE)
 

package/arka/skills/costs/SKILL.md

@@ -70,10 +70,10 @@ Outside this skill's scope but worth surfacing alongside the summary:
 | Lever | Where | Effect |
 | --- | --- | --- |
 | `ARKAOS_LLM_PROVIDER` | env | Switch between Anthropic, OpenAI, Gemini adapters per the multi-LLM router. |
-| `fallbackModel` | `~/.claude/settings.json` (Claude Code 2.1.166+) | Up to three models tried in order when the primary is overloaded or unavailable. `npx arkaos install` / `update` seed `["claude-opus-5", "claude-sonnet-5"]` when the key is absent and never overwrite an operator chain; `/arka status` shows the configured chain. |
-| Schedule `model:` / `fallback_models:` | `~/.arkaos/schedules.yaml` | A pinned `model` goes to `claude --model`; without one the Model Fabric `strategy` role becomes `ANTHROPIC_DEFAULT_MODEL` (Claude Code 2.1.236+). `fallback_models` goes to `--fallback-model` (default `claude-opus-5 → claude-sonnet-5`) so one overload or model 404 no longer ends a nightly cycle. |
+| `fallbackModel` | `~/.claude/settings.json` (Claude Code 2.1.166+) | Up to three models tried in order when the primary is overloaded or unavailable. `npx arkaos install` / `update` seed `["claude-opus-5-5", "claude-sonnet-5"]` when the key is absent, upgrade a chain that is exactly a previous ArkaOS default, and never overwrite an operator chain; `/arka status` shows the configured chain. |
+| Schedule `model:` / `fallback_models:` | `~/.arkaos/schedules.yaml` | A pinned `model` goes to `claude --model`; without one the Model Fabric `strategy` role becomes `ANTHROPIC_DEFAULT_MODEL` (Claude Code 2.1.236+). `fallback_models` goes to `--fallback-model` (default `claude-opus-5-5 → claude-sonnet-5`) so one overload or model 404 no longer ends a nightly cycle. |
 | `ANTHROPIC_BEDROCK_SERVICE_TIER` | env (Claude Code 2.1.122+) | `default` / `flex` / `priority`. Only relevant when routing through AWS Bedrock — `flex` cuts cost ~50 % on non-urgent workloads at the price of higher tail latency. |
-| Agent `model:` field | `departments/*/agents/*.yaml` | Per-agent override per the model-routing matrix in CLAUDE.md. Mechanical roles default to `sonnet` at low effort, C-suite to `opus` (Opus 5); the Quality Gate row below lifts Marta / Eduardo / Francisca to the frontier tier. |
+| Agent `model:` field | `departments/*/agents/*.yaml` | Per-agent override per the model-routing matrix in CLAUDE.md. Mechanical roles default to `sonnet` at low effort, C-suite to `opus` (Opus 5.5); the Quality Gate row below lifts Marta / Eduardo / Francisca to the frontier tier. |
 | Quality Gate model | constitution `quality_gate.model_policy` | Marta / Eduardo / Francisca run on the best model available (frontier tier) regardless of the cost ceiling — review quality is never the place to save tokens. |
 
 ## Non-negotiables

package/bin/arka-doctor

@@ -370,17 +370,23 @@ run_check "plugins" "warn" "$PLUGINS_DESC" "$PLUGINS_OK" "Run: bash install.sh"
 
 # ─── Check 17: Hyperframes Skills (video production) ──────────────────────
 # Parity with installer/doctor.js `hyperframes-skills` (PR-C2): same name,
-# same sentinel, warn-only — video production is opt-in.
+# same sentinels, same two roots (~/.claude/skills and ~/.agents/skills),
+# warn-only — video production is opt-in.
 HF_OK=false
 HF_DESC="Hyperframes skills AUSENTE (video-as-code editing for /content video)"
-for hf_name in hyperframes hyperframes-core; do
-  if [ -f "$HOME/.claude/skills/$hf_name/SKILL.md" ]; then
-    HF_OK=true
-    HF_DESC="Hyperframes skills installed ($hf_name)"
-    break
-  fi
+for hf_root in "$HOME/.claude/skills" "$HOME/.agents/skills"; do
+  for hf_name in hyperframes hyperframes-core; do
+    if [ -f "$hf_root/$hf_name/SKILL.md" ]; then
+      HF_OK=true
+      # Static, matching installer/doctor.js `description` exactly: run_check
+      # splices $desc unescaped into the --json payload, so no path or
+      # name interpolation may reach it.
+      HF_DESC="Hyperframes skills installed (video-as-code editing for /content video)"
+      break 2
+    fi
+  done
 done
-run_check "hyperframes-skills" "warn" "$HF_DESC" "$HF_OK" "Run /content video-setup in Claude Code"
+run_check "hyperframes-skills" "warn" "$HF_DESC" "$HF_OK" "Run /content video-setup in Claude Code, or: npx hyperframes skills update (core set; npx hyperframes skills check reports staleness)"
 
 # ─── Check 18: SQLite corrupt-store backups (self-heal F1-D1) ──────────────
 # Parity with installer/doctor.js `sqlite-corrupt-backups`: same name,

package/config/cognition/schedules.yaml

@@ -5,10 +5,10 @@
 #
 # Model pin and fallback chain (Runtime Sync PR3) — per schedule, optional:
 #   model: claude-sonnet-5            # pinned: goes to `claude --model`
-#   fallback_models: [claude-opus-5]  # `--fallback-model a,b`; [] disables
+#   fallback_models: [claude-opus-5-5]  # `--fallback-model a,b`; [] disables
 # Without `model`, the Model Fabric `strategy` role (~/.arkaos/models.yaml)
 # becomes ANTHROPIC_DEFAULT_MODEL for the run. Without `fallback_models`,
-# the chain is claude-opus-5 → claude-sonnet-5. The runtime caps chains at
+# the chain is claude-opus-5-5 → claude-sonnet-5. The runtime caps chains at
 # three models after de-duplication, removes ids outside `availableModels`
 # before the walk and fails over past an id it cannot reach (a retired
 # model); the ArkaOS gateway plays no part in the chain.

package/core/harness/drift.py

@@ -462,25 +462,36 @@ def _check_worktree(report: DriftReport, settings: dict[str, Any]) -> None:
         )
 
 
+def _fallback_finding(status: DriftStatus, detail: str) -> DriftFinding:
+    return DriftFinding("settings:fallbackModel", status, "fallbackModel", detail)
+
+
 def _check_fallback_model(report: DriftReport, settings: dict[str, Any]) -> None:
     """MISSING when the key is absent or JSON null (the seeder writes the
-    default in both cases); ADOPTED for any other value that is not the
-    default — an array, an empty array, or the legacy string form."""
-    from core.runtime.claude_code import DEFAULT_FALLBACK_MODELS
+    default in both cases); DIVERGED when the chain is exactly a default
+    ArkaOS seeded in an earlier release (assert upgrades it); ADOPTED for
+    any other value that is not the default — an array, an empty array,
+    or the legacy string form."""
+    from core.runtime.claude_code import (
+        DEFAULT_FALLBACK_MODELS,
+        is_previous_fallback_default,
+    )
 
     chain = settings.get("fallbackModel")
     if chain is None:
         report.findings.append(
-            DriftFinding(
-                "settings:fallbackModel", DriftStatus.MISSING, "fallbackModel",
-                "fallbackModel chain default not seeded",
-            )
+            _fallback_finding(DriftStatus.MISSING, "fallbackModel chain default not seeded")
+        )
+    elif is_previous_fallback_default(chain):
+        detail = (
+            f"previous ArkaOS default chain ({' → '.join(chain)}); "
+            f"`arka harness assert` upgrades it to {' → '.join(DEFAULT_FALLBACK_MODELS)}"
         )
+        report.findings.append(_fallback_finding(DriftStatus.DIVERGED, detail))
     elif chain != list(DEFAULT_FALLBACK_MODELS):
         report.findings.append(
-            DriftFinding(
-                "settings:fallbackModel", DriftStatus.ADOPTED, "fallbackModel",
-                "operator-configured fallback chain; seed policy adopts it",
+            _fallback_finding(
+                DriftStatus.ADOPTED, "operator-configured fallback chain; seed policy adopts it"
             )
         )
 

package/core/harness/manager.py

@@ -41,7 +41,10 @@ from pathlib import Path
 from core.harness import drift, json_store, paths
 from core.harness.manifest import OwnershipManifest
 from core.harness.spec import HookRegistration, spec_for
-from core.runtime.claude_code import DEFAULT_FALLBACK_MODELS
+from core.runtime.claude_code import (
+    DEFAULT_FALLBACK_MODELS,
+    is_previous_fallback_default,
+)
 
 
 @dataclass
@@ -49,8 +52,8 @@ class SurfaceAction:
     """One thing assert did (or deliberately did not do) to a surface."""
 
     surface: str
-    # created | repaired | unrepaired | reseeded | adopted-skip
-    # | refused | noop
+    # created | repaired | unrepaired | reseeded | upgraded
+    # | adopted-skip | refused | noop
     action: str
     detail: str = ""
 
@@ -353,13 +356,23 @@ class ClaudeConfigManager:
     def _assert_seed(
         self, settings: dict, report: AssertReport, reseed: bool
     ) -> None:
-        for surface, key, default, is_ours in _seed_surfaces(self):
+        for surface, key, default, is_ours, is_previous in _seed_surfaces(self):
             current = settings.get(key)
             if current is None:
                 settings[key] = default
                 report.actions.append(SurfaceAction(surface, "created"))
             elif is_ours(current):
                 report.actions.append(SurfaceAction(surface, "noop"))
+            elif is_previous(current):
+                # A chain ArkaOS itself seeded before is not an operator
+                # decision: move it to the current default.
+                settings[key] = default
+                report.actions.append(
+                    SurfaceAction(
+                        surface, "upgraded",
+                        "previous ArkaOS default; moved to the current default",
+                    )
+                )
             elif reseed:
                 settings[key] = default
                 report.actions.append(SurfaceAction(surface, "reseeded"))
@@ -531,20 +544,27 @@ def _user_deny_extensions(home: Path | None) -> list[str]:
     return [r for r in raw if isinstance(r, str) and r]
 
 
+def _never_previous(_value: object) -> bool:
+    """Surfaces with no history of ArkaOS-authored defaults to upgrade."""
+    return False
+
+
 def _seed_surfaces(manager: ClaudeConfigManager):
     statusline = _statusline_default(manager)
     return (
         (
             "settings:statusLine", "statusLine", statusline,
-            drift._is_arkaos_statusline,
+            drift._is_arkaos_statusline, _never_previous,
         ),
         (
             "settings:worktree", "worktree", {"baseRef": "head"},
             lambda v: isinstance(v, dict) and v.get("baseRef") == "head",
+            _never_previous,
         ),
         (
             "settings:fallbackModel", "fallbackModel", list(DEFAULT_FALLBACK_MODELS),
             lambda v: v == list(DEFAULT_FALLBACK_MODELS),
+            is_previous_fallback_default,
         ),
     )
 

package/core/hooks/session_start.py

@@ -318,31 +318,6 @@ def _trigger_reorganizer(repo: str, config: dict) -> None:
     _spawn_detached([sys.executable, "-m", "core.cognition.reorganizer_cli"], repo)
 
 
-def _trigger_auto_doc_drain(repo: str, config: dict) -> None:
-    """Queue-aware AND config-gated: ``cognition.auto_doc_drain_on_session``
-    (default True). The Stop hook enqueues auto-doc jobs; sessions drain
-    them — the consumer follows the producer's cadence, so a machine with
-    no sessions never wakes a worker for nothing (knowledge-flow PR1)."""
-    cognition = config.get("cognition") or {}
-    if not cognition.get("auto_doc_drain_on_session", True):
-        return
-    if not repo:
-        return
-    try:
-        from core.jobs.auto_doc_worker import _queue_root
-        if not any((_queue_root() / "pending").glob("*.json")):
-            return
-    except Exception:
-        return
-    max_age = cognition.get("auto_doc_max_age_days", 14)
-    _spawn_detached(
-        [sys.executable, "-m", "core.jobs.auto_doc_worker", "--once",
-         "--max-jobs", "10", "--max-age-days", str(max_age)],
-        repo,
-        log_path=Path.home() / ".arkaos" / "logs" / "auto-doc-drain.log",
-    )
-
-
 def _ensure_dashboard(repo: str, config: dict) -> None:
     dashboard = config.get("dashboard") or {}
     if not dashboard.get("ensure_on_session", True):
@@ -409,7 +384,6 @@ def build_visible(cwd: str) -> str:
     msg += _forge_line()
     msg += _drift(version)
     _trigger_reorganizer(repo, config)
-    _trigger_auto_doc_drain(repo, config)
     _ensure_dashboard(repo, config)
     return msg
 

package/core/jobs/auto_doc_worker.py

@@ -6,8 +6,6 @@ Queue layout under `~/.arkaos/jobs/auto-doc/`:
     processing/  — jobs currently being handled
     completed/   — jobs finished successfully
     failed/      — jobs that exhausted retries or could not be parsed
-    expired/     — jobs older than --max-age-days, moved aside unprocessed
-                   (never deleted: the operator can audit or re-queue them)
 
 Each job is a JSON file:
 
@@ -28,31 +26,22 @@ second worker that tries to claim the same job will raise and skip it.
 from __future__ import annotations
 
 import argparse
-import contextlib
 import json
 import os
 import sys
 import time
 import uuid
-from datetime import UTC, datetime
+from datetime import datetime, timezone
 from pathlib import Path
+from typing import Optional
 
-from core.knowledge.vault import resolve_vault_path
 from core.shared import safe_session_id as _safe_session_id_module
 
+
 MAX_ATTEMPTS = 3
 # Re-export for backward compatibility with any external importers.
 SAFE_SESSION_ID_RE = _safe_session_id_module.SAFE_SESSION_ID_RE
-_QUEUE_SUBDIRS = ("pending", "processing", "completed", "failed", "expired")
-
-_VAULT_REMEDIATION = (
-    "No vault configured. Set `knowledge.vaultPath` in ~/.arkaos/config.json "
-    "(or export ARKAOS_VAULT). Jobs were left untouched in pending/."
-)
-
-
-class VaultNotConfiguredError(RuntimeError):
-    """No vault answered; pending jobs must not be claimed or retried."""
+_QUEUE_SUBDIRS = ("pending", "processing", "completed", "failed")
 
 
 def _queue_root() -> Path:
@@ -62,17 +51,11 @@ def _queue_root() -> Path:
     return Path.home() / ".arkaos" / "jobs" / "auto-doc"
 
 
-def _vault_path() -> Path | None:
-    """The operator's vault via the canonical resolver, or ``None``.
-
-    Resolution order is the canonical one from ``core.knowledge.vault``:
-    ``knowledge.vaultPath`` in ``~/.arkaos/config.json`` first, then the
-    ``ARKAOS_VAULT`` env var, both required to exist on disk. This
-    replaces a hardcoded ``~/Documents/Personal`` fallback that silently
-    sent every note to one developer's personal layout (the exact defect
-    ``core.knowledge.vault`` was written to end — the worker predates it).
-    """
-    return resolve_vault_path()
+def _vault_path() -> Path:
+    override = os.environ.get("ARKAOS_VAULT", "").strip()
+    if override:
+        return Path(override)
+    return Path.home() / "Documents" / "Personal"
 
 
 def _ensure_queue(root: Path) -> None:
@@ -81,7 +64,7 @@ def _ensure_queue(root: Path) -> None:
 
 
 def _now_iso() -> str:
-    return datetime.now(UTC).isoformat()
+    return datetime.now(timezone.utc).isoformat()
 
 
 # ─── Enqueue ───────────────────────────────────────────────────────────
@@ -91,7 +74,7 @@ def enqueue_job(
     session_id: str,
     transcript_path: str,
     qg_verdict: str,
-    queue_root: Path | None = None,
+    queue_root: Optional[Path] = None,
 ) -> str:
     """Write a pending job file. Returns the job id."""
     root = queue_root or _queue_root()
@@ -118,71 +101,22 @@ def enqueue_job(
 
 def process_pending_jobs(
     max_jobs: int = 10,
-    queue_root: Path | None = None,
-    max_age_days: float | None = None,
+    queue_root: Optional[Path] = None,
 ) -> list[dict]:
-    """Process up to `max_jobs` pending jobs. Returns per-job result dicts.
-
-    Raises :class:`VaultNotConfiguredError` BEFORE claiming anything when
-    no vault is configured — a missing vault is an environment failure
-    and must never consume job attempts or move files out of pending/.
-    """
+    """Process up to `max_jobs` pending jobs. Returns per-job result dicts."""
     root = queue_root or _queue_root()
     _ensure_queue(root)
-    vault = _vault_path()
-    if vault is None:
-        raise VaultNotConfiguredError(_VAULT_REMEDIATION)
-    pending = sorted((root / "pending").glob("*.json"))
     results: list[dict] = []
-    if max_age_days is not None:
-        pending, expired = _partition_expired(pending, root, max_age_days)
-        results.extend(expired)
+    pending = sorted((root / "pending").glob("*.json"))
     for job_file in pending[:max_jobs]:
         claimed = _claim(job_file, root)
         if claimed is None:
             continue
-        results.append(run_single_job(claimed, root, vault=vault))
+        results.append(run_single_job(claimed, root))
     return results
 
 
-def _job_timestamp(job_file: Path) -> float | None:
-    prefix = job_file.name.split("-", 1)[0]
-    try:
-        return float(prefix)
-    except ValueError:
-        return None
-
-
-def _partition_expired(
-    pending: list[Path], root: Path, max_age_days: float
-) -> tuple[list[Path], list[dict]]:
-    """Split by job age, read from the unix prefix of the job filename.
-
-    A name without a parsable prefix is kept for processing — unknown
-    age must never silently shelve a job.
-    """
-    cutoff = time.time() - max_age_days * 86400.0
-    keep: list[Path] = []
-    expired: list[dict] = []
-    for job_file in pending:
-        stamp = _job_timestamp(job_file)
-        if stamp is not None and stamp < cutoff:
-            expired.append(_move_to_expired(job_file, root))
-        else:
-            keep.append(job_file)
-    return keep, expired
-
-
-def _move_to_expired(job_file: Path, root: Path) -> dict:
-    target = root / "expired" / job_file.name
-    try:
-        os.replace(job_file, target)
-    except OSError:
-        return {"job_id": job_file.stem, "status": "expire-failed"}
-    return {"job_id": job_file.stem, "status": "expired"}
-
-
-def _claim(job_file: Path, root: Path) -> Path | None:
+def _claim(job_file: Path, root: Path) -> Optional[Path]:
     target = root / "processing" / job_file.name
     try:
         os.replace(job_file, target)
@@ -193,17 +127,10 @@ def _claim(job_file: Path, root: Path) -> Path | None:
         return None
 
 
-def run_single_job(
-    job_path: Path,
-    queue_root: Path | None = None,
-    vault: Path | None = None,
-) -> dict:
+def run_single_job(job_path: Path, queue_root: Optional[Path] = None) -> dict:
     """Process one job end-to-end. Moves to completed/ or failed/."""
     root = queue_root or _queue_root()
     _ensure_queue(root)
-    vault = vault or _vault_path()
-    if vault is None:
-        return _release_claim(job_path, root)
     try:
         payload = json.loads(job_path.read_text(encoding="utf-8"))
         if not isinstance(payload, dict):
@@ -213,26 +140,18 @@ def run_single_job(
     payload["attempts"] = int(payload.get("attempts") or 0) + 1
     job_path.write_text(json.dumps(payload), encoding="utf-8")
     try:
-        written = _execute(payload, vault)
+        written = _execute(payload)
     except Exception as exc:
         return _handle_retry(job_path, payload, root, str(exc))
     return _move_to_completed(job_path, root, payload, written)
 
 
-def _release_claim(job_path: Path, root: Path) -> dict:
-    """Vault vanished between pre-flight and execution: put the job back
-    WITHOUT touching attempts — environment failure, not job failure."""
-    target = root / "pending" / job_path.name
-    with contextlib.suppress(OSError):
-        os.replace(job_path, target)
-    return {"job_id": job_path.stem, "status": "vault-unconfigured"}
-
-
-def _execute(payload: dict, vault: Path) -> list[str]:
+def _execute(payload: dict) -> list[str]:
     from core.cognition.auto_documentor import document_session
     transcript = Path(payload.get("transcript_path") or "")
     session_id = str(payload.get("session_id") or "")
     verdict = str(payload.get("qg_verdict") or "")
+    vault = _vault_path()
     paths = document_session(transcript, session_id, vault, verdict)
     return [str(p) for p in paths]
 
@@ -264,8 +183,10 @@ def _move_to_completed(
     payload["status"] = "completed"
     target = root / "completed" / job_path.name
     job_path.write_text(json.dumps(payload), encoding="utf-8")
-    with contextlib.suppress(OSError):
+    try:
         os.replace(job_path, target)
+    except OSError:
+        pass
     return {
         "job_id": payload.get("job_id"),
         "status": "completed",
@@ -279,13 +200,17 @@ def _move_to_failed(job_path: Path, root: Path, error: str) -> dict:
     info["status"] = "failed"
     info["error"] = error
     info["failed_at"] = _now_iso()
-    with contextlib.suppress(OSError):
+    try:
         job_path.write_text(json.dumps(info), encoding="utf-8")
+    except OSError:
+        pass
     try:
         os.replace(job_path, target)
     except OSError:
-        with contextlib.suppress(OSError):
+        try:
             job_path.unlink()
+        except OSError:
+            pass
     return {
         "job_id": info.get("job_id"),
         "status": "failed",
@@ -314,38 +239,20 @@ def _build_parser() -> argparse.ArgumentParser:
                         help="Cap on jobs per pass.")
     parser.add_argument("--interval", type=float, default=5.0,
                         help="Seconds to sleep between passes when daemonised.")
-    parser.add_argument("--max-age-days", type=float, default=None,
-                        help="Move pending jobs older than this to expired/ "
-                             "instead of processing them.")
     return parser
 
 
-def _run_pass(max_jobs: int, max_age_days: float | None) -> int | None:
-    """One drain pass; prints results. Returns an exit code on fatal error."""
-    try:
-        results = process_pending_jobs(
-            max_jobs=max_jobs, max_age_days=max_age_days
-        )
-    except VaultNotConfiguredError as exc:
-        print(str(exc), file=sys.stderr)
-        return 2
+def main(argv: Optional[list[str]] = None) -> int:
+    args = _build_parser().parse_args(argv)
+    results = process_pending_jobs(max_jobs=args.max_jobs)
     for r in results:
         print(json.dumps(r), flush=True)
-    return None
-
-
-def main(argv: list[str] | None = None) -> int:
-    args = _build_parser().parse_args(argv)
-    rc = _run_pass(args.max_jobs, args.max_age_days)
-    if rc is not None:
-        return rc
     if args.once:
         return 0
     while True:
         time.sleep(max(0.1, float(args.interval)))
-        rc = _run_pass(args.max_jobs, args.max_age_days)
-        if rc is not None:
-            return rc
+        for r in process_pending_jobs(max_jobs=args.max_jobs):
+            print(json.dumps(r), flush=True)
 
 
 if __name__ == "__main__":

package/core/runtime/claude_code.py

@@ -108,7 +108,19 @@ _PROBE_TIMEOUT_SECONDS = 10
 # model is overloaded or unavailable; the runtime caps chains at three after
 # de-duplication. installer/fallback-model.js carries the same list —
 # tests/python/test_scheduler_daemon.py pins the two to each other.
-DEFAULT_FALLBACK_MODELS: tuple[str, ...] = ("claude-opus-5", "claude-sonnet-5")
+DEFAULT_FALLBACK_MODELS: tuple[str, ...] = ("claude-opus-5-5", "claude-sonnet-5")
+# Chains ArkaOS ITSELF seeded in earlier releases. A settings.json holding one
+# of these byte for byte is not an operator decision, so the seeder and
+# `arka harness assert` upgrade it to DEFAULT_FALLBACK_MODELS; any other
+# present value is the operator's and stays. Append, never edit — the JS twin
+# in installer/fallback-model.js is pinned to this list by the same test.
+PREVIOUS_FALLBACK_DEFAULTS: tuple[tuple[str, ...], ...] = (("claude-opus-5", "claude-sonnet-5"),)
+
+
+def is_previous_fallback_default(chain: object) -> bool:
+    """True when ``chain`` is exactly a chain ArkaOS seeded in the past."""
+    return isinstance(chain, list) and tuple(chain) in PREVIOUS_FALLBACK_DEFAULTS
+
 
 # Cache for the `claude --version` probe, keyed by the binary probed ("" =
 # the PATH lookup). In-memory only: a disk cache would let one session

package/core/runtime/gateway/litellm_config.py

@@ -51,10 +51,11 @@ _SLOT_ROLE: dict[str, str] = {
 
 # runtime-tier tokens -> real Anthropic model ids the proxy calls upstream.
 # Real ids (claude-fable-5-1, claude-*) pass straight through.
-# Runtime Sync 2026-09-03: `best` is Fable 5.1, `opus` is Opus 5, and the
-# weakest lane (`haiku`/`fast`) is Sonnet 5 — ArkaOS never routes to Haiku.
+# Runtime Sync 2026-09-03 (+ Opus 5.5 sweep 2026-09-22): `best` is Fable 5.1,
+# `opus` is Opus 5.5 (the id the Claude Code 2.1.280 alias resolves to), and
+# the weakest lane (`haiku`/`fast`) is Sonnet 5 — ArkaOS never routes to Haiku.
 _RUNTIME_TO_ANTHROPIC: dict[str, str] = {
-    "opus": "claude-opus-5",
+    "opus": "claude-opus-5-5",
     "best": "claude-fable-5-1",
     "fable": "claude-fable-5-1",
     "sonnet": "claude-sonnet-5",

package/core/runtime/model_router.py

@@ -55,18 +55,23 @@ KNOWN_EFFORTS = ("low", "medium", "high", "max")
 
 _ALIAS_NAMES = frozenset({"best", "default", "fast"})
 
-# Runtime Sync 2026-09-03 — ids that no longer exist, or lanes ArkaOS no
-# longer routes to (Opus 4.x → Opus 5, Fable 5 → Fable 5.1, Haiku → Sonnet 5).
-# Normalised when a role is RESOLVED; the operator's models.yaml is never
-# rewritten. One stderr notice per process per id, so a stale pin is
-# visible without becoming noise.
+# Runtime Sync 2026-09-03 (Opus 4.x → Opus 5, Fable 5 → Fable 5.1, Haiku →
+# Sonnet 5) and the Opus 5.5 sweep 2026-09-22 (Opus 5 → Opus 5.5) — ids that
+# no longer exist, or lanes ArkaOS no longer routes to. Normalised when a
+# role is RESOLVED; the operator's models.yaml is never rewritten. One
+# stderr notice per process per id, so a stale pin is visible without
+# becoming noise. Legacy ids map straight to the CURRENT lane — the lookup
+# is a single dict read, not a transitive walk, so no value may be a key.
+LEGACY_SYNC_DATE = "2026-09-22"
 LEGACY_MODEL_IDS: dict[str, str] = {
     "claude-fable-5": "claude-fable-5-1",
     "claude-fable-5[1m]": "claude-fable-5-1[1m]",
-    "claude-opus-4-8": "claude-opus-5",
-    "claude-opus-4-8[1m]": "claude-opus-5[1m]",
-    "claude-opus-4-7": "claude-opus-5",
-    "claude-opus-4-6": "claude-opus-5",
+    "claude-opus-5": "claude-opus-5-5",
+    "claude-opus-5[1m]": "claude-opus-5-5[1m]",
+    "claude-opus-4-8": "claude-opus-5-5",
+    "claude-opus-4-8[1m]": "claude-opus-5-5[1m]",
+    "claude-opus-4-7": "claude-opus-5-5",
+    "claude-opus-4-6": "claude-opus-5-5",
     "claude-haiku-4-5-20251001": "claude-sonnet-5",
     "claude-haiku-4-5": "claude-sonnet-5",
     "haiku": "sonnet",
@@ -89,7 +94,8 @@ def normalise_model_id(model: str, source: str = "models.yaml") -> str:
         _LEGACY_NOTICED.add((source, model))
         print(
             f"[arka:warn] {source} pins legacy model {model!r}; resolving as "
-            f"{current!r} (Runtime Sync 2026-09-03). Update the pin to silence this.",
+            f"{current!r} (retired lane; last sweep {LEGACY_SYNC_DATE}). "
+            "Update the pin to silence this.",
             file=sys.stderr,
         )
     return current