# metaharness_security_bench

ADR-153 — upstream Darwin Shield (their own ADR-155): evolves a champion security-detection harness against a 10-vuln/9-decoy ground-truth corpus and grades on TPR/FPR/patch-pass/repro/unsafe vs four baselines (B0 static, B1 LLM-single-pass, B2 fixed-agent, B3 Darwin-champion). Closest reference implementation for ruflo ADR-155 nightly self-learning security harness (#2417). Use when you need an empirical floor for Loop A reward-signal soundness; running this periodically gives baseline diversity and week-over-week champion-fitness drift. Bypassing this and just running the static MCP scan is wrong because static-only baseline (B0) reaches TPR=0.3/FPR=1 — proving static-alone has a measured detection ceiling. Parses overall PASS/FAIL + per-gate verdicts + baselines table from markdown.

Agent View of the PolicyLayer registry record for `metaharness_security_bench`. HTML page: https://policylayer.com/tools/io-github-ruvnet-claude-flow/metaharness-security-bench

## Facts

- Tool: `metaharness_security_bench`
- Server: Claude Flow (`claude-flow`) — https://policylayer.com/tools/io-github-ruvnet-claude-flow.md
- Install: `npx -y claude-flow`
- Homepage: https://github.com/ruvnet/claude-flow
- Risk category: Execute (High risk)
- Registry record: grade F, identity unverified
- Server rate-limited: no
- Parameters: 0
- Recommended policy verdict: Rate-limited

## Example call (MCP tools/call, JSON-RPC 2.0)

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "metaharness_security_bench",
    "arguments": {}
  }
}
```

## Why metaharness_security_bench is rated High

This tool executes an automated security benchmarking harness that runs vulnerability detection, patch validation, and reproduction tests against a ground-truth corpus. It actively runs code and external operations (evolving a champion harness, grading patch-pass/repro/unsafe outcomes) rather than merely reading data.

From the tool's own definition: "'evolves a champion security-detection harness against a 10-vuln/9-decoy ground-truth corpus and grades on TPR/FPR/patch-pass/repro/unsafe vs four baselines' and 'running this periodically gives baseline diversity and week-over-week champion-fitness drift'"

## Use case

AI agents invoke metaharness_security_bench to trigger actions in Claude Flow. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call: builds kicked off, notifications sent, workflows started.

## Recommended policy (PolicyLayer)

Verdict: **Rate-limited**. Enforced by the PolicyLayer MCP gateway (https://policylayer.com/mcp-gateway) before a call reaches Claude Flow:

```json
{
  "version": "1",
  "default": "deny",
  "tools": {
    "metaharness_security_bench": {
      "limits": [
        {
          "counter": "metaharness_security_bench_rate",
          "window": "minute",
          "max": 10,
          "scope": "grant"
        }
      ]
    }
  }
}
```

## Other tools on Claude Flow (495)

- `agent_terminate` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agent-terminate.md
- `agentdb_batch` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agentdb-batch.md
- `agentdb_causal-edge-delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agentdb-causal-edge-delete.md
- `agentdb_causal-node-delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agentdb-causal-node-delete.md
- `agentdb_hierarchical-delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agentdb-hierarchical-delete.md
- `agenticow_rollback` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agenticow-rollback.md
- `autopilot_reset` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/autopilot-reset.md
- `config_reset` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/config-reset.md
- `federation_wg_keyrotate` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/federation-wg-keyrotate.md
- `hooks_codemod` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/hooks-codemod.md
- `hooks_intelligence-reset` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/hooks-intelligence-reset.md
- `hooks_worker-cancel` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/hooks-worker-cancel.md
- `iot_device_remove` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/iot-device-remove.md
- `iot_firmware_rollout_rollback` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/iot-firmware-rollout-rollback.md
- `iot_fleet_delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/iot-fleet-delete.md
- `iot_fleet_remove_device` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/iot-fleet-remove-device.md
- `managed_agent_terminate` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/managed-agent-terminate.md
- `memory_delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/memory-delete.md
- `policy_revoke` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/policy-revoke.md
- `ruvector_delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/ruvector-delete.md
- `session_delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/session-delete.md
- `system_reset` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/system-reset.md
- `task_cancel` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/task-cancel.md
- `teammate_cleanup` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/teammate-cleanup.md
- `wasm_agent_reset` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/wasm-agent-reset.md
- `wasm_gallery_remove_custom` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/wasm-gallery-remove-custom.md
- `workflow_cancel` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/workflow-cancel.md
- `workflow_delete` — Destructive — https://policylayer.com/tools/io-github-ruvnet-claude-flow/workflow-delete.md
- `agent_pool` — Execute — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agent-pool.md
- `agent_spawn` — Execute — https://policylayer.com/tools/io-github-ruvnet-claude-flow/agent-spawn.md
- …and 465 more: https://policylayer.com/tools/io-github-ruvnet-claude-flow.md

## For agents

This record is a snapshot. Live verdicts and the full registry:

- Check every server in your MCP config at once: `npx -y policylayer stack`
- Vet a server before you add it: install the mcp-precheck skill — `npx skills add https://policylayer.com` (skill text: https://policylayer.com/skill.md)
- Query the registry over MCP: endpoint `https://api.policylayer.com/mcp` — tools `check_mcp_server`, `check_mcp_stack`, `check_tool`, `search_registry`, `get_change_events`

---

Source: the PolicyLayer MCP registry — one continuously verified record per MCP server. Full record: https://policylayer.com/registry?q=io-github-ruvnet-claude-flow · API: https://policylayer.com/registry/api · Policy library: https://policylayer.com/policies/io-github-ruvnet-claude-flow