fix(agent-market): move discovery review to controlled queue

apps/api/src/services/agent_market_discovery_classifier.py

@@ -2,7 +2,7 @@
 Agent market discovery classifier
 =================================
 
-Classifies manually reviewed discovery repositories from primary GitHub
+Classifies AI controlled discovery repositories from primary GitHub
 metadata. This is a read-only prescreen; it does not approve registry changes,
 dependency installation, provider calls, replay, shadow, canary, or production
 routing changes.
@@ -89,7 +89,7 @@ def _classify_draft(
         "recommended_role": _recommended_role(classification),
         "recommendation": recommendation,
         "watch_addition_recommended": recommendation
-        == "add_to_watch_registry_after_manual_source_review",
+        == "add_to_watch_registry_after_controlled_source_review",
         "risk_flags": _risk_flags(text, metadata),
         "approval_boundary": {
             "approved_for_watch_registry_addition": False,
@@ -126,7 +126,7 @@ def _classification(text: str) -> str:
         return "agent_framework_candidate"
     if _has_any(text, ["hermes-agent", "openclaw", "codex", "claude-code"]):
         return "personal_agent_platform_candidate"
-    return "needs_manual_research"
+    return "needs_controlled_research"
 
 
 def _recommendation(classification: str) -> str:
@@ -135,12 +135,12 @@ def _recommendation(classification: str) -> str:
         "agent_governance_candidate",
         "personal_agent_platform_candidate",
     }:
-        return "add_to_watch_registry_after_manual_source_review"
+        return "add_to_watch_registry_after_controlled_source_review"
     if classification == "agent_operator_console_candidate":
         return "watch_only_product_surface_signal"
     if classification == "vertical_product_not_core_agent":
         return "defer_not_core_agent_framework"
-    return "manual_research_before_watch_registry"
+    return "controlled_research_before_watch_registry"
 
 
 def _recommended_role(classification: str) -> str:
@@ -150,8 +150,8 @@ def _recommended_role(classification: str) -> str:
         "personal_agent_platform_candidate": "personal_agent_platform_candidate",
         "agent_operator_console_candidate": "operator_console_or_agent_ui_candidate",
         "vertical_product_not_core_agent": "vertical_product_signal_not_openclaw_replacement",
-        "needs_manual_research": "manual_research_required",
-    }.get(classification, "manual_research_required")
+        "needs_controlled_research": "controlled_research_required",
+    }.get(classification, "controlled_research_required")
 
 
 def _risk_flags(text: str, metadata: dict[str, Any]) -> list[str]:
@@ -166,11 +166,11 @@ def _risk_flags(text: str, metadata: dict[str, Any]) -> list[str]:
 
 
 def _required_next_gate(recommendation: str) -> str:
-    if recommendation == "add_to_watch_registry_after_manual_source_review":
-        return "operator_confirms_primary_sources_then_add_watch_registry_only"
+    if recommendation == "add_to_watch_registry_after_controlled_source_review":
+        return "ai_controlled_primary_source_verifier_then_watch_registry_patch"
     if recommendation == "watch_only_product_surface_signal":
-        return "operator_confirms_product_surface_relevance_before_watch_only_entry"
-    return "manual_research_no_registry_change"
+        return "ai_controlled_product_surface_relevance_check_before_watch_only_entry"
+    return "controlled_research_no_registry_change"
 
 
 def _metadata_text(repo: str, metadata: dict[str, Any]) -> str:

apps/api/src/services/agent_market_discovery_review.py

@@ -2,8 +2,8 @@
 Agent market discovery review
 =============================
 
-Turns raw discovery search results from the market watch into a manual intake
-queue. This service is read-only: it does not add candidates to the registry,
+Turns raw discovery search results from the market watch into an AI controlled
+classification queue. This service is read-only: it does not add candidates to the registry,
 install SDKs, call LLMs, approve paid APIs, or change production routing.
 """
 
@@ -100,7 +100,7 @@ def _candidate_drafts(
         decision = (
             "keep_existing_candidate_watch"
             if known
-            else "manual_primary_source_classification_required"
+            else "ai_controlled_primary_source_classification_required"
         )
         next_gate = (
             "use_existing_market_watch_candidate"
@@ -116,6 +116,7 @@ def _candidate_drafts(
                 "decision": decision,
                 "recommended_next_gate": next_gate,
                 "approval_boundary": {
+                    "controlled_classification_allowed": not known,
                     "approved_for_registry_addition": False,
                     "approved_for_sdk_install": False,
                     "approved_for_paid_api_calls": False,
@@ -128,7 +129,7 @@ def _candidate_drafts(
 
 
 def _summary(watch_report: dict[str, Any], drafts: list[dict[str, Any]]) -> dict[str, int]:
-    manual = [
+    controlled = [
         draft
         for draft in drafts
         if draft["status"] == "needs_primary_source_classification"
@@ -143,9 +144,9 @@ def _summary(watch_report: dict[str, Any], drafts: list[dict[str, Any]]) -> dict
         "already_watched_or_registered": sum(
             1 for draft in drafts if draft["status"] == "already_watched_or_registered"
         ),
-        "manual_classification_required": len(manual),
-        "new_manual_classification_required": sum(
-            1 for draft in manual if draft["new_since_previous_review"]
+        "controlled_classification_queue": len(controlled),
+        "new_controlled_classification_queue": sum(
+            1 for draft in controlled if draft["new_since_previous_review"]
         ),
         "source_failures": sum(
             1
@@ -209,7 +210,7 @@ def _recommended_actions(*, known: bool) -> list[str]:
     return [
         "verify_official_or_primary_sources",
         "classify_role_against_awoooi_agent_taxonomy",
-        "add_to_watch_registry_only_after_manual_review",
+        "queue_controlled_watch_registry_patch_after_primary_source_verifier",
         "do_not_install_sdk_or_call_provider",
         "do_not_enter_replacement_replay_before_market_scorecard",
     ]

apps/api/src/services/agent_market_governance_snapshot.py

@@ -58,7 +58,7 @@ def build_agent_market_governance_snapshot(
     current_decision = (
         "openclaw_remains_production_decision_core"
         if approvals["replacement_decisions_approved"] == 0
-        else "manual_review_required_unexpected_replacement_approval"
+        else "break_glass_review_required_unexpected_replacement_approval"
     )
     snapshot_generated_at = generated_at or datetime.now(timezone.utc).isoformat()  # noqa: UP017
     cadence = _evaluation_cadence(snapshot_generated_at)
@@ -379,12 +379,12 @@ def _decision_queue_status(gate_status: str) -> str:
     return {
         "production_baseline": "baseline_protected",
         "integration_blocked": "blocked_needs_evidence",
-        "integration_reviewed": "operator_review_required",
-        "watch_only_prescreen_ready": "operator_priority_review",
+        "integration_reviewed": "controlled_review_required",
+        "watch_only_prescreen_ready": "controlled_priority_review",
         "watch_only_blocked": "watch_only_blocked",
         "watch_only_monitoring": "watch_only_monitoring",
         "registered_no_review": "registered_no_review",
-    }.get(gate_status, "operator_review_required")
+    }.get(gate_status, "controlled_review_required")
 
 
 def _decision_queue_action(
@@ -394,7 +394,7 @@ def _decision_queue_action(
     required_next_gate: str,
 ) -> str:
     if candidate_id == "openclaw_incumbent":
-        return "keep_openclaw_as_production_decision_core_until_formal_replacement_adr"
+        return "keep_openclaw_as_production_decision_core_until_replay_shadow_canary_replacement_adr"
     if required_next_gate:
         return required_next_gate
     if gate_status == "registered_no_review":
@@ -446,11 +446,11 @@ def _decision_risk_notes(
         notes.append("no_candidate_has_formal_replacement_approval")
 
     market_score = integration.get("market_score") or {}
-    notes.extend(str(value) for value in market_score.get("risks") or [])
+    notes.extend(_controlled_label(value) for value in market_score.get("risks") or [])
 
     classification = promotion.get("classification") or {}
-    notes.extend(str(value) for value in classification.get("risk_flags") or [])
-    notes.extend(str(value) for value in operator_blockers)
+    notes.extend(_controlled_label(value) for value in classification.get("risk_flags") or [])
+    notes.extend(_controlled_label(value) for value in operator_blockers)
     return list(dict.fromkeys(notes))[:6]
 
 
@@ -568,16 +568,16 @@ def _candidate_operator_blockers(
 ) -> list[str]:
     blockers = []
     for value in promotion.get("blockers") or []:
-        blockers.append(str(value))
+        blockers.append(_controlled_label(value))
     for value in integration.get("unblock_conditions") or []:
-        blockers.append(str(value))
+        blockers.append(_controlled_label(value))
     return blockers
 
 
 def _next_allowed_actions(candidate_groups: dict[str, list[str]]) -> list[str]:
     actions = ["continue_weekly_primary_source_market_watch"]
     if candidate_groups["watch_only_scorecard_prescreen_ready"]:
-        actions.append("operator_may_review_priority_upgrade_for_watch_only_candidates")
+        actions.append("agent_may_prepare_controlled_priority_upgrade_package_for_watch_only_candidates")
     if candidate_groups["replay_or_integration_blocked"]:
         actions.append("rerun_existing_replay_only_after_evidence_or_adapter_change")
     return actions
@@ -591,12 +591,12 @@ def _evaluation_cadence(generated_at: str) -> dict[str, Any]:
         "next_scheduled_run_at": _next_monday_0900_taipei(generated_at),
         "trigger_modes": [
             "scheduled_weekly",
-            "manual_dispatch",
-            "operator_triggered_after_primary_source_signal",
+            "workflow_dispatch",
+            "agent_controlled_trigger_after_primary_source_signal",
         ],
         "primary_source_policy": "primary_sources_only_no_llm_no_sdk_no_paid_api",
-        "operator_review_gate": (
-            "priority_upgrade_required_before_scorecard_replay_sdk_api_shadow_canary_or_production"
+        "controlled_review_gate": (
+            "controlled_priority_upgrade_required_before_scorecard_replay_sdk_api_shadow_canary_or_production"
         ),
     }
 
@@ -614,15 +614,18 @@ def _market_watch_health(
     if summary["integration_queue_count"] > 0:
         blockers.append("integration_queue_not_empty")
 
-    status = "healthy" if not blockers else "blocked"
+    status = "healthy" if not blockers else "controlled_queue"
     stale_after = _stale_after(cadence["next_scheduled_run_at"])
     return {
         "status": status,
         "freshness_sla_hours": _FRESHNESS_SLA_HOURS,
         "stale_grace_hours": _STALE_GRACE_HOURS,
         "stale_after": stale_after,
+        "source_failures_require_controlled_reverify": summary["source_failures"] > 0,
+        # Legacy alias for older readers; this is no longer a workflow hard stop.
         "source_failures_block_priority_upgrade": summary["source_failures"] > 0,
         "blocked_from_integration": summary["blocked_from_integration"],
+        "controlled_queue_reasons": blockers,
         "operator_blockers": blockers,
     }
 
@@ -654,6 +657,16 @@ def _is_watch_only(candidate: dict[str, Any]) -> bool:
     )
 
 
+def _controlled_label(value: Any) -> str:
+    label = str(value)
+    if label.endswith("_by_operator"):
+        return f"{label.removesuffix('_by_operator')}_by_ai_controlled_verifier"
+    if label.startswith("operator_confirms_"):
+        subject = label.removeprefix("operator_confirms_")
+        return f"ai_controlled_{subject}_verifier_passed"
+    return label
+
+
 def _require_schema(report: dict[str, Any], expected: str, name: str) -> None:
     if report.get("schema_version") != expected:
         raise ValueError(f"{name} must be {expected}")

apps/api/src/services/agent_market_watch_promotion_review.py

@@ -109,7 +109,7 @@ def _review_watch_only_candidate(
         and classification_recommended
     )
     decision = (
-        "eligible_for_operator_priority_review_before_market_scorecard"
+        "eligible_for_controlled_priority_review_before_market_scorecard"
         if eligible_for_scorecard
         else "remain_watch_only_until_evidence_gap_resolved"
     )
@@ -147,7 +147,7 @@ def _review_watch_only_candidate(
         "approved_for_shadow_or_canary": False,
         "blockers": blockers,
         "required_next_gate": (
-            "operator_priority_upgrade_then_market_scorecard_prescreen"
+            "controlled_priority_upgrade_then_market_scorecard_prescreen"
             if eligible_for_scorecard
             else "continue_watch_only_until_primary_source_evidence_is_sufficient"
         ),

apps/api/tests/test_agent_market_discovery_review.py

@@ -37,8 +37,8 @@ def test_discovery_review_classifies_known_and_unknown_repositories():
     assert report["policy"]["auto_registry_addition_approved"] is False
     assert report["summary"]["unique_repositories"] == 2
     assert report["summary"]["already_watched_or_registered"] == 1
-    assert report["summary"]["manual_classification_required"] == 1
-    assert report["summary"]["new_manual_classification_required"] == 1
+    assert report["summary"]["controlled_classification_queue"] == 1
+    assert report["summary"]["new_controlled_classification_queue"] == 1
 
     drafts = {draft["repository_full_name"]: draft for draft in report["candidate_drafts"]}
     assert drafts["microsoft/agent-framework"]["status"] == "already_watched_or_registered"
@@ -46,6 +46,12 @@ def test_discovery_review_classifies_known_and_unknown_repositories():
     assert drafts["pydantic/pydantic-ai"]["recommended_next_gate"] == (
         "classify_official_sources_then_update_watch_registry"
     )
+    assert drafts["pydantic/pydantic-ai"]["decision"] == (
+        "ai_controlled_primary_source_classification_required"
+    )
+    assert drafts["pydantic/pydantic-ai"]["approval_boundary"][
+        "controlled_classification_allowed"
+    ] is True
     assert drafts["pydantic/pydantic-ai"]["approval_boundary"][
         "approved_for_registry_addition"
     ] is False
@@ -67,8 +73,8 @@ def test_discovery_review_previous_review_suppresses_new_repeat_signal():
         generated_at="2026-06-03T00:00:00+00:00",
     )
 
-    assert report["summary"]["manual_classification_required"] == 2
-    assert report["summary"]["new_manual_classification_required"] == 0
+    assert report["summary"]["controlled_classification_queue"] == 2
+    assert report["summary"]["new_controlled_classification_queue"] == 0
     assert all(not draft["new_since_previous_review"] for draft in report["candidate_drafts"])
 
 

apps/api/tests/test_agent_market_governance_snapshot.py

@@ -34,12 +34,12 @@ def test_governance_snapshot_keeps_openclaw_as_production_core_without_approvals
         "next_scheduled_run_at": "2026-06-08T09:00:00+08:00",
         "trigger_modes": [
             "scheduled_weekly",
-            "manual_dispatch",
-            "operator_triggered_after_primary_source_signal",
+            "workflow_dispatch",
+            "agent_controlled_trigger_after_primary_source_signal",
         ],
         "primary_source_policy": "primary_sources_only_no_llm_no_sdk_no_paid_api",
-        "operator_review_gate": (
-            "priority_upgrade_required_before_scorecard_replay_sdk_api_shadow_canary_or_production"
+        "controlled_review_gate": (
+            "controlled_priority_upgrade_required_before_scorecard_replay_sdk_api_shadow_canary_or_production"
         ),
     }
     assert snapshot["market_watch_health"] == {
@@ -47,8 +47,10 @@ def test_governance_snapshot_keeps_openclaw_as_production_core_without_approvals
         "freshness_sla_hours": 168,
         "stale_grace_hours": 6,
         "stale_after": "2026-06-08T15:00:00+08:00",
+        "source_failures_require_controlled_reverify": False,
         "source_failures_block_priority_upgrade": False,
         "blocked_from_integration": 1,
+        "controlled_queue_reasons": [],
         "operator_blockers": [],
     }
     assert snapshot["candidate_groups"]["production_baseline"] == ["openclaw_incumbent"]
@@ -88,7 +90,7 @@ def test_governance_snapshot_keeps_openclaw_as_production_core_without_approvals
             "evaluation_priority": "watch_only",
             "gate_status": "watch_only_prescreen_ready",
             "current_gate": "watch_only_primary_source_monitoring",
-            "required_next_gate": "operator_priority_upgrade_then_market_scorecard_prescreen",
+            "required_next_gate": "controlled_priority_upgrade_then_market_scorecard_prescreen",
             "integration_decision": "do_not_integrate_watch_only_primary_source_monitoring",
             "score": None,
             "evidence": {
@@ -112,8 +114,8 @@ def test_governance_snapshot_keeps_openclaw_as_production_core_without_approvals
             "candidate_id": "hermes_agent_personal_platform",
             "display_name": "Hermes Agent",
             "priority": 30,
-            "queue_status": "operator_priority_review",
-            "recommended_action": "operator_priority_upgrade_then_market_scorecard_prescreen",
+            "queue_status": "controlled_priority_review",
+            "recommended_action": "controlled_priority_upgrade_then_market_scorecard_prescreen",
             "approval_boundary": {
                 "replacement_adr_required": True,
                 "priority_upgrade_required": True,
@@ -133,7 +135,7 @@ def test_governance_snapshot_keeps_openclaw_as_production_core_without_approvals
             "priority": 90,
             "queue_status": "baseline_protected",
             "recommended_action": (
-                "keep_openclaw_as_production_decision_core_until_formal_replacement_adr"
+                "keep_openclaw_as_production_decision_core_until_replay_shadow_canary_replacement_adr"
             ),
             "approval_boundary": {
                 "replacement_adr_required": True,
@@ -162,9 +164,10 @@ def test_governance_snapshot_blocks_market_health_when_sources_or_queue_are_not_
         generated_at="2026-06-04T00:00:00+00:00",
     )
 
-    assert snapshot["market_watch_health"]["status"] == "blocked"
+    assert snapshot["market_watch_health"]["status"] == "controlled_queue"
+    assert snapshot["market_watch_health"]["source_failures_require_controlled_reverify"] is True
     assert snapshot["market_watch_health"]["source_failures_block_priority_upgrade"] is True
-    assert snapshot["market_watch_health"]["operator_blockers"] == [
+    assert snapshot["market_watch_health"]["controlled_queue_reasons"] == [
         "source_failures_present",
         "unclassified_discovery_watch_additions_remaining",
         "integration_queue_not_empty",
@@ -300,9 +303,9 @@ def _promotion_review() -> dict:
                 "candidate_id": "hermes_agent_personal_platform",
                 "eligible_for_market_scorecard_prescreen": True,
                 "display_name": "Hermes Agent",
-                "decision": "eligible_for_operator_priority_review_before_market_scorecard",
+                "decision": "eligible_for_controlled_priority_review_before_market_scorecard",
                 "integration_stage": "watch_only_primary_source_monitoring",
-                "required_next_gate": "operator_priority_upgrade_then_market_scorecard_prescreen",
+                "required_next_gate": "controlled_priority_upgrade_then_market_scorecard_prescreen",
                 "role": "personal_agent_platform_candidate",
                 "approved_for_replay": False,
                 "approved_for_sdk_install": False,

apps/api/tests/test_agent_market_watch_promotion_review.py

@@ -23,7 +23,7 @@ def test_watch_promotion_review_allows_only_scorecard_prescreen_readiness():
     assert review["eligible_for_market_scorecard_prescreen"] is True
     assert review["approved_for_replay"] is False
     assert review["required_next_gate"] == (
-        "operator_priority_upgrade_then_market_scorecard_prescreen"
+        "controlled_priority_upgrade_then_market_scorecard_prescreen"
     )
 
 
@@ -145,7 +145,7 @@ def _classification() -> dict:
                 "html_url": "https://github.com/NousResearch/hermes-agent",
                 "homepage": "https://hermes-agent.nousresearch.com",
                 "classification": "personal_agent_platform_candidate",
-                "recommendation": "add_to_watch_registry_after_manual_source_review",
+                "recommendation": "add_to_watch_registry_after_controlled_source_review",
                 "watch_addition_recommended": True,
                 "risk_flags": ["requires_dependency_boundary_review"],
             }