{
  "candidate": "manual-hardcase-repair-v3-cardinality-max4096",
  "concurrency": 2,
  "harness": "localpager-agent",
  "mean_score": 0.6233333333333333,
  "offset": 0,
  "routing_policy_path": "prompt-optimizer/out/manual-hardcase-repair-20260614T092140Z/routing_policy.md",
  "routing_policy_sha256": "b2576ca027148e109a1e72029c192ea9e26be486508671ec7c11025ac80f948b",
  "row_reports": [
    {
      "error": null,
      "false_negatives": [
        "agent_runtime"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "gateway",
        "agent_runtime"
      ],
      "id": "openclaw-openclaw-48940",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "gateway"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/48940",
      "title": "ACP: add gateway-owned node-backed runtime",
      "true_positives": [
        "acp",
        "gateway"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "security"
      ],
      "false_positives": [],
      "gold_topics": [
        "mcp_tooling",
        "config",
        "security"
      ],
      "id": "openclaw-openclaw-80783",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "config",
        "mcp_tooling"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/80783",
      "title": "Policy: add model, network, and MCP conformance checks",
      "true_positives": [
        "config",
        "mcp_tooling"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "cron_automation"
      ],
      "false_positives": [],
      "gold_topics": [
        "exec_tools",
        "browser_automation",
        "cron_automation"
      ],
      "id": "openclaw-openclaw-42027",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "exec_tools",
        "browser_automation"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/42027",
      "title": "fix: resolve exec PATH fallback, layered browser diagnostics, and cron force-run deadlock",
      "true_positives": [
        "exec_tools",
        "browser_automation"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "codex",
        "chat_integrations"
      ],
      "id": "openclaw-openclaw-77748",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "codex",
        "chat_integrations"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/77748",
      "title": "fix: Codex startup plugins + WhatsApp history & Docker Codex OAuth",
      "true_positives": [
        "codex",
        "chat_integrations"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "model_serving"
      ],
      "id": "openclaw-openclaw-79897",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "model_serving"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/79897",
      "title": "OpenAI-compatible streaming with llama.cpp saves zero usage (stream closed before final usage chunk)",
      "true_positives": [
        "model_serving"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "acp"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "approvals",
        "acpx"
      ],
      "id": "openclaw-openclaw-40332",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "approvals",
        "acpx"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/40332",
      "title": "[Feature]: Per-binding and per-agent permissionMode for ACP sessions",
      "true_positives": [
        "approvals",
        "acpx"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "gateway",
        "sessions"
      ],
      "id": "openclaw-openclaw-63007",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "gateway",
        "sessions"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/63007",
      "title": "Pass outbound session identity into message_sending and surface guarded gateway send denial",
      "true_positives": [
        "gateway",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "memory",
        "reliability"
      ],
      "id": "openclaw-openclaw-80255",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "memory",
        "reliability"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/80255",
      "title": "fix #79026: active-memory recall subagent can deadlock on the main lane inside before_prompt_build",
      "true_positives": [
        "memory",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "api_surface"
      ],
      "false_positives": [],
      "gold_topics": [
        "gateway",
        "api_surface",
        "ui_tui"
      ],
      "id": "openclaw-openclaw-84670",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "gateway",
        "ui_tui"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/84670",
      "title": "[codex] fix webchat full-message reader for truncated history",
      "true_positives": [
        "gateway",
        "ui_tui"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "queueing",
        "docs"
      ],
      "id": "openclaw-openclaw-46552",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "docs",
        "queueing"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/46552",
      "title": "docs(queue): clarify steer behavior with partial streaming and tool boundaries",
      "true_positives": [
        "docs",
        "queueing"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "approvals"
      ],
      "false_positives": [],
      "gold_topics": [
        "exec_tools",
        "sandboxing",
        "approvals"
      ],
      "id": "openclaw-openclaw-62428",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "exec_tools",
        "sandboxing"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/62428",
      "title": "test(exec): land exec v2 contract follow-through",
      "true_positives": [
        "exec_tools",
        "sandboxing"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "codex"
      ],
      "false_positives": [],
      "gold_topics": [
        "acpx",
        "codex",
        "skills_plugins"
      ],
      "id": "openclaw-openclaw-82507",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acpx",
        "skills_plugins"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/82507",
      "title": "[Feature]: ACPX Codex sandbox should inherit user-installed plugins (e.g. Superpowers)",
      "true_positives": [
        "acpx",
        "skills_plugins"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "self_hosted_inference",
        "memory"
      ],
      "id": "openclaw-openclaw-80479",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "memory",
        "self_hosted_inference"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/80479",
      "title": "feat(memory/embeddings): add openai-compatible provider for self-hosted servers (llama.cpp, Ollama, vLLM, TGI, LocalAI)",
      "true_positives": [
        "memory",
        "self_hosted_inference"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [
        "model_serving"
      ],
      "gold_topics": [
        "local_model_providers",
        "reliability"
      ],
      "id": "openclaw-openclaw-90146",
      "loss": 3.0,
      "over_label_count": 0,
      "predicted_topics": [
        "local_model_providers",
        "model_serving"
      ],
      "score": 0.25,
      "target": "https://github.com/openclaw/openclaw/issues/90146",
      "title": "google-vertex: Missing gemini-3.1-flash-lite in provider catalog causes silent failure instead of error",
      "true_positives": [
        "local_model_providers"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "docs"
      ],
      "id": "openclaw-openclaw-51849",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "docs"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/51849",
      "title": "Docs: add freeCodeCamp OpenClaw full tutorial to showcase",
      "true_positives": [
        "docs"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "local_model_providers"
      ],
      "false_positives": [],
      "gold_topics": [
        "open_weight_models",
        "local_model_providers"
      ],
      "id": "openclaw-openclaw-68725",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "open_weight_models"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/68725",
      "title": "feat(amazon-bedrock-mantle): add known context windows for open-weight Mantle models",
      "true_positives": [
        "open_weight_models"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "notifications",
        "chat_integrations"
      ],
      "id": "openclaw-openclaw-84297",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "notifications",
        "chat_integrations"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/84297",
      "title": "[Bug]: Per-agent identity overlay dropped on cron --announce and heartbeat target-channel Slack pushes (announce path; reply path was fixed in #38235)",
      "true_positives": [
        "notifications",
        "chat_integrations"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "model_serving"
      ],
      "false_positives": [],
      "gold_topics": [
        "model_serving",
        "local_models"
      ],
      "id": "openclaw-openclaw-77827",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "local_models"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/77827",
      "title": "fix: LM Studio thinking blocks invisible with Responses API",
      "true_positives": [
        "local_models"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "security"
      ],
      "false_positives": [
        "packaging_deployment"
      ],
      "gold_topics": [
        "security"
      ],
      "id": "openclaw-openclaw-81957",
      "loss": 3.0,
      "over_label_count": 0,
      "predicted_topics": [
        "packaging_deployment"
      ],
      "score": 0.25,
      "target": "https://github.com/openclaw/openclaw/pull/81957",
      "title": "ci: harden GitHub Actions supply-chain boundaries",
      "true_positives": []
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents"
      ],
      "false_positives": [],
      "gold_topics": [
        "coding_agents",
        "sandboxing",
        "agent_runtime"
      ],
      "id": "openclaw-openclaw-39248",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sandboxing",
        "agent_runtime"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/39248",
      "title": "Bug: sandbox.mode: \"non-main\" silently breaks sessions_spawn subagent initialization",
      "true_positives": [
        "sandboxing",
        "agent_runtime"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "sessions",
        "telemetry_usage"
      ],
      "false_positives": [
        "ui_tui"
      ],
      "gold_topics": [
        "sessions",
        "telemetry_usage"
      ],
      "id": "openclaw-openclaw-47083",
      "loss": 4.0,
      "over_label_count": 0,
      "predicted_topics": [
        "ui_tui"
      ],
      "score": 0.2,
      "target": "https://github.com/openclaw/openclaw/pull/47083",
      "title": "fix: respect totalTokensFresh flag to avoid showing stale token counts",
      "true_positives": []
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "mcp_tooling",
        "tool_calling"
      ],
      "id": "openclaw-openclaw-70882",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "mcp_tooling",
        "tool_calling"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/70882",
      "title": "fix(bundle-mcp): coerce stringified object/array params before MCP tool calls",
      "true_positives": [
        "mcp_tooling",
        "tool_calling"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "hooks"
      ],
      "false_positives": [],
      "gold_topics": [
        "security",
        "hooks",
        "skills_plugins"
      ],
      "id": "openclaw-openclaw-63826",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "security",
        "skills_plugins"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/63826",
      "title": "security: fix HIGH/CRITICAL vulns in skill scanner, SSRF, hook priority, and token verification",
      "true_positives": [
        "security",
        "skills_plugins"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "local_models",
        "self_hosted_inference"
      ],
      "id": "openclaw-openclaw-81249",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "local_models",
        "self_hosted_inference"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/81249",
      "title": "[Feature/Bug]: Local Ollama embeddings fail when proxy is enabled (SSRF defenses ignore NO_PROXY)",
      "true_positives": [
        "local_models",
        "self_hosted_inference"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "packaging_deployment"
      ],
      "false_positives": [],
      "gold_topics": [
        "browser_automation",
        "packaging_deployment"
      ],
      "id": "openclaw-openclaw-70529",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "browser_automation"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/70529",
      "title": "[Bug]: Desktop cannot use existing Chrome sessions: EasyClaw Google sign-in fails, and user profile attach fails with spawn npx ENOENT",
      "true_positives": [
        "browser_automation"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "local_model_providers",
        "model_serving"
      ],
      "id": "openclaw-openclaw-87277",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "local_model_providers",
        "model_serving"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/87277",
      "title": "[Feature] Add MiMo-V2.5 to Xiaomi catalog + automatic multimodal routing when DeepSeek V4-Pro is primary model",
      "true_positives": [
        "local_model_providers",
        "model_serving"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "sessions"
      ],
      "id": "openclaw-openclaw-64199",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "sessions"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/64199",
      "title": "[Bug]: ACP configured binding uses parent channel ID for session key \u2014 all threads under same channel share one persistent Claude Code process",
      "true_positives": [
        "acp",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "sessions"
      ],
      "false_positives": [],
      "gold_topics": [
        "reliability",
        "auth_identity",
        "sessions"
      ],
      "id": "openclaw-openclaw-84752",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "reliability",
        "auth_identity"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/84752",
      "title": "fix: self-heal lane wedges + restore openai-codex OAuth on embedded path",
      "true_positives": [
        "reliability",
        "auth_identity"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [],
      "gold_topics": [
        "cron_automation",
        "sessions",
        "reliability"
      ],
      "id": "openclaw-openclaw-84583",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "cron_automation",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/84583",
      "title": "cron announce delivery triggers EmbeddedAttemptSessionTakeoverError when user is actively chatting",
      "true_positives": [
        "cron_automation",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "acpx",
        "acp"
      ],
      "id": "openclaw-openclaw-67244",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "acpx"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/67244",
      "title": "Explicit ACP agent runs: embedded backend visibility failure and stale final JSON state after sessions_yield",
      "true_positives": [
        "acp",
        "acpx"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "sandboxing"
      ],
      "false_positives": [],
      "gold_topics": [
        "config",
        "sandboxing",
        "gateway"
      ],
      "id": "openclaw-openclaw-71216",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "config",
        "gateway"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/71216",
      "title": "Config schema: add `sandbox`, `routing.rules`, `instances`, and `gateway.nodes.denyPaths`",
      "true_positives": [
        "config",
        "gateway"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "agent_runtime"
      ],
      "false_positives": [],
      "gold_topics": [
        "sessions",
        "agent_runtime",
        "reliability"
      ],
      "id": "openclaw-openclaw-84477",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sessions",
        "reliability"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/84477",
      "title": "Discord embedded-run prep wedge before strict-agentic, recovery skips sessionId=unknown lanes",
      "true_positives": [
        "sessions",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "coding_agents",
        "reliability"
      ],
      "id": "openclaw-openclaw-65242",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "reliability"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/65242",
      "title": "fix: CompletionDeliveryGate to prevent duplicate ACP completion delivery",
      "true_positives": [
        "acp",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "acpx",
        "auth_identity"
      ],
      "false_positives": [],
      "gold_topics": [
        "codex",
        "acp",
        "acpx",
        "auth_identity"
      ],
      "id": "openclaw-openclaw-73910",
      "loss": 2.0,
      "over_label_count": 0,
      "predicted_topics": [
        "codex",
        "acp"
      ],
      "score": 0.3333333333333333,
      "target": "https://github.com/openclaw/openclaw/issues/73910",
      "title": "BUG: OpenClaw-managed Codex ACP uses isolated CODEX_HOME without auth bridge and sends unsupported timeout config",
      "true_positives": [
        "codex",
        "acp"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents"
      ],
      "false_positives": [
        "skills_plugins"
      ],
      "gold_topics": [
        "acp",
        "coding_agents"
      ],
      "id": "openclaw-openclaw-80008",
      "loss": 3.0,
      "over_label_count": 0,
      "predicted_topics": [
        "skills_plugins",
        "acp"
      ],
      "score": 0.25,
      "target": "https://github.com/openclaw/openclaw/pull/80008",
      "title": "feat(plugins): expose ACP spawn and prompt in plugin runtime",
      "true_positives": [
        "acp"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "cron_automation"
      ],
      "false_positives": [],
      "gold_topics": [
        "reliability",
        "exec_tools",
        "cron_automation"
      ],
      "id": "openclaw-openclaw-43765",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "reliability",
        "exec_tools"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/43765",
      "title": "Improve runtime recovery for heartbeat, Feishu, and exec sessions",
      "true_positives": [
        "reliability",
        "exec_tools"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "chat_integrations"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "chat_integrations",
        "sessions"
      ],
      "id": "openclaw-openclaw-60979",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/60979",
      "title": "feature: sessions_spawn ACP delivery to channel (stream output to Zulip/Discord topic)",
      "true_positives": [
        "acp",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "agent_runtime"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "codex",
        "agent_runtime"
      ],
      "id": "openclaw-openclaw-83863",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "codex"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/83863",
      "title": "ACP/Codex child tasks can be marked succeeded with progress-only output and no final deliverable",
      "true_positives": [
        "acp",
        "codex"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "packaging_deployment"
      ],
      "false_positives": [],
      "gold_topics": [
        "codex",
        "packaging_deployment"
      ],
      "id": "openclaw-openclaw-84715",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "codex"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/84715",
      "title": "[Bug]: @openclaw/codex peer link failure reproduced on 2026.5.19 after update",
      "true_positives": [
        "codex"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [],
      "gold_topics": [
        "sessions",
        "chat_integrations",
        "reliability"
      ],
      "id": "openclaw-openclaw-84757",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sessions",
        "chat_integrations"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/84757",
      "title": "[Bug]: Telegram session can get stuck after compaction when encrypted reasoning content fails verification",
      "true_positives": [
        "sessions",
        "chat_integrations"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "agent_runtime"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "sessions",
        "agent_runtime"
      ],
      "id": "openclaw-openclaw-56442",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/56442",
      "title": "feat: Add opt-in ACP parent completion notify for sessions_spawn",
      "true_positives": [
        "acp",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "skills_plugins"
      ],
      "false_positives": [],
      "gold_topics": [
        "security",
        "exec_tools",
        "skills_plugins"
      ],
      "id": "openclaw-openclaw-78528",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "security",
        "exec_tools"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/78528",
      "title": "Security: skill SecretRef API keys still leak into exec child environments",
      "true_positives": [
        "security",
        "exec_tools"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "memory",
        "sessions"
      ],
      "id": "openclaw-openclaw-84789",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "memory",
        "sessions"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/84789",
      "title": "Active memory crashes on Telegram forum topic sessions (dirName validation)",
      "true_positives": [
        "memory",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "security"
      ],
      "false_positives": [],
      "gold_topics": [
        "acpx",
        "acp",
        "security"
      ],
      "id": "openclaw-openclaw-84763",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "acpx"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/84763",
      "title": "fix(acpx): scrub provider credential env from ACP harness spawns",
      "true_positives": [
        "acp",
        "acpx"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "api_surface"
      ],
      "false_positives": [
        "skills_plugins"
      ],
      "gold_topics": [
        "auth_identity",
        "api_surface"
      ],
      "id": "openclaw-openclaw-65364",
      "loss": 3.0,
      "over_label_count": 0,
      "predicted_topics": [
        "skills_plugins",
        "auth_identity"
      ],
      "score": 0.25,
      "target": "https://github.com/openclaw/openclaw/pull/65364",
      "title": "feat(plugins): add registerProviderRuntimeAuthOverride API",
      "true_positives": [
        "auth_identity"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "sessions",
        "reliability"
      ],
      "id": "openclaw-openclaw-52747",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/52747",
      "title": "fix(acp): time out stuck session lane tasks",
      "true_positives": [
        "acp",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents"
      ],
      "false_positives": [],
      "gold_topics": [
        "queueing",
        "sessions",
        "coding_agents"
      ],
      "id": "openclaw-openclaw-10467",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "queueing",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/10467",
      "title": "[Feature Request]: Multi-lane concurrency support for sub-agents via sessions_spawn",
      "true_positives": [
        "queueing",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "tool_calling",
        "security"
      ],
      "id": "openclaw-openclaw-43246",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "security",
        "tool_calling"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/pull/43246",
      "title": "fix(message): deny same-provider cross-context sends by default [AI-assisted]",
      "true_positives": [
        "security",
        "tool_calling"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "sessions",
        "reliability"
      ],
      "id": "openclaw-openclaw-59878",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sessions",
        "reliability"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/59878",
      "title": "Session lane stuck in 'running' after run dies \u2014 sessions.abort + gateway restart fail to clear stale state",
      "true_positives": [
        "sessions",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "security"
      ],
      "false_positives": [],
      "gold_topics": [
        "model_serving",
        "security",
        "config"
      ],
      "id": "openclaw-openclaw-51667",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "model_serving",
        "config"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/51667",
      "title": "Feature: Native Audio Input for Omni-Modal Models (skip STT transcription)",
      "true_positives": [
        "model_serving",
        "config"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "self_hosted_inference"
      ],
      "false_positives": [],
      "gold_topics": [
        "local_models",
        "memory",
        "self_hosted_inference"
      ],
      "id": "openclaw-openclaw-44202",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "local_models",
        "memory"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/44202",
      "title": "[Bug]: local memory embeddings on Apple Silicon can crash gateway in ggml-metal / node-llama-cpp; need official Metal/GPU guidance",
      "true_positives": [
        "local_models",
        "memory"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "sessions"
      ],
      "false_positives": [],
      "gold_topics": [
        "acpx",
        "codex",
        "sessions"
      ],
      "id": "openclaw-openclaw-48580",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acpx",
        "codex"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/48580",
      "title": "Bug: acpx codex sessions \u521b\u5efa\u7684\u4f1a\u8bdd\u7acb\u5373\u9000\u51fa - stdin is not a terminal",
      "true_positives": [
        "acpx",
        "codex"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "acpx",
        "codex"
      ],
      "id": "openclaw-openclaw-74305",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acpx",
        "codex"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/74305",
      "title": "[Bug]: ACPX Codex worker fails when model/thinking overrides are configured",
      "true_positives": [
        "acpx",
        "codex"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents",
        "reliability"
      ],
      "false_positives": [
        "agent_runtime"
      ],
      "gold_topics": [
        "tool_calling",
        "coding_agents",
        "reliability"
      ],
      "id": "openclaw-openclaw-45393",
      "loss": 4.0,
      "over_label_count": 0,
      "predicted_topics": [
        "tool_calling",
        "agent_runtime"
      ],
      "score": 0.2,
      "target": "https://github.com/openclaw/openclaw/pull/45393",
      "title": "fix(errors): friendly message and last-message repair for tool_use/tool_result mismatch (#45385)",
      "true_positives": [
        "tool_calling"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "reliability",
        "sessions"
      ],
      "id": "openclaw-openclaw-84771",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sessions",
        "reliability"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/84771",
      "title": "Event loop saturation during startup: synchronous model-prewarm and session-locks block event loop for 28-64 seconds",
      "true_positives": [
        "sessions",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "coding_agents",
        "memory",
        "hooks"
      ],
      "false_positives": [
        "agent_runtime"
      ],
      "gold_topics": [
        "coding_agents",
        "memory",
        "hooks",
        "reliability"
      ],
      "id": "openclaw-openclaw-44379",
      "loss": 5.0,
      "over_label_count": 0,
      "predicted_topics": [
        "agent_runtime",
        "reliability"
      ],
      "score": 0.16666666666666666,
      "target": "https://github.com/openclaw/openclaw/pull/44379",
      "title": "fix(pi-runner): harden context-overflow recovery with one suppress-hook retry",
      "true_positives": [
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [],
      "false_positives": [],
      "gold_topics": [
        "reliability",
        "sessions"
      ],
      "id": "openclaw-openclaw-84746",
      "loss": 0.0,
      "over_label_count": 0,
      "predicted_topics": [
        "sessions",
        "reliability"
      ],
      "score": 1.0,
      "target": "https://github.com/openclaw/openclaw/issues/84746",
      "title": "[Bug]: Auto-compaction crashes active responses after 5.18 transcript lock scope change (#13744)",
      "true_positives": [
        "sessions",
        "reliability"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "gateway"
      ],
      "false_positives": [],
      "gold_topics": [
        "mcp_tooling",
        "sessions",
        "gateway"
      ],
      "id": "openclaw-openclaw-68187",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "mcp_tooling",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/68187",
      "title": "SSE-backed MCP sessions can stay stale after server restart and fail with 'Session not found'",
      "true_positives": [
        "mcp_tooling",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [],
      "gold_topics": [
        "acp",
        "sessions",
        "reliability"
      ],
      "id": "openclaw-openclaw-52249",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "acp",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/issues/52249",
      "title": "ACP parent session stuck until refresh when yielded waiting for child completion",
      "true_positives": [
        "acp",
        "sessions"
      ]
    },
    {
      "error": null,
      "false_negatives": [
        "reliability"
      ],
      "false_positives": [],
      "gold_topics": [
        "cron_automation",
        "sessions",
        "reliability"
      ],
      "id": "openclaw-openclaw-69256",
      "loss": 1.0,
      "over_label_count": 0,
      "predicted_topics": [
        "cron_automation",
        "sessions"
      ],
      "score": 0.5,
      "target": "https://github.com/openclaw/openclaw/pull/69256",
      "title": "fix(cron): prevent premature session cleanup when subagents are running",
      "true_positives": [
        "cron_automation",
        "sessions"
      ]
    }
  ],
  "rows": 60,
  "scores": [
    0.5,
    0.5,
    0.5,
    1.0,
    1.0,
    0.5,
    1.0,
    1.0,
    0.5,
    1.0,
    0.5,
    0.5,
    1.0,
    0.25,
    1.0,
    0.5,
    1.0,
    0.5,
    0.25,
    0.5,
    0.2,
    1.0,
    0.5,
    1.0,
    0.5,
    1.0,
    1.0,
    0.5,
    0.5,
    1.0,
    0.5,
    0.5,
    0.5,
    0.3333333333333333,
    0.25,
    0.5,
    0.5,
    0.5,
    0.5,
    0.5,
    0.5,
    0.5,
    1.0,
    0.5,
    0.25,
    0.5,
    0.5,
    1.0,
    1.0,
    0.5,
    0.5,
    0.5,
    1.0,
    0.2,
    1.0,
    0.16666666666666666,
    1.0,
    0.5,
    0.5,
    0.5
  ]
}