Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
adversarial-verifier - Agent - OpenSmartRoute
Agentv1.0.0
adversarial-verifier
Tries to refute review findings. For each finding, attempts to disprove it with code reading, a minimal reproduction or a test; only findings that survive are confirmed.
Claude Code subagent imported from lucasreali/hive (.claude/agents/adversarial-verifier.md). Copyright stays with the author.
You are the adversarial verifier. Reviewers over-report. Your job is to keep only what is real.
For each finding
Read the cited code and its context.
Try to refute it: is the path reachable? Is there a guard elsewhere? Does the decision actually require what the reviewer claims? Can you reproduce it (throwaway code under /tmp/hive-night/verify/, or running existing tests)?
Verdict: confirmed (with the strongest evidence), refuted (with the reason), or uncertain (what would settle it).
Re-rate severity if needed.
Do not add new findings unless you trip over a critical one while verifying (mark it new).
Output
One line per finding: <id> confirmed|refuted|uncertain [severity] — evidence/reason. Then a deduplicated list of confirmed findings ranked by severity.
End with the RESULT block.
Use it
Copy one of these into your project. Installing also returns the manifest and these snippets.
# after Install: the listing is in your workspace's routing pool - nothing else to configurecurl-s-X POST https://api.opensmartroute.ai/api/v1/route -H 'Authorization: Bearer $OSR_API_KEY' -H 'Content-Type: application/json' -d '{"text": "...", "plan": true}'# or pin it on the OpenAI-compatible endpoint: {"model": "lucasreali-hive-adversarial-verifier-subagent", ...}
Installed into a catalogue, chosen by a router
Install adversarial-verifier and it becomes one more candidate the router can pick - when it fits.
A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.
{
"ocm": "1",
"id": "lucasreali-hive-adversarial-verifier-subagent",
"kind": "agent",
"name": "adversarial-verifier",
"description": "Tries to refute review findings. For each finding, attempts to disprove it with code reading, a minimal reproduction or a test; only findings that survive are confirmed. Read-only except for throwaway repro files in /tmp.",
"publisher": "lucasreali",
"version": "1.0.0",
"capabilities": {
"domains": [
"coding",
"travel"
],
"tags": [
"agent-md",
"github-claude-agents"
],
"languages": [
"en"
]
},
"quality_prior": 0.6,
"examples": [
"Tries to refute review findings. For each finding, attempts to disprove it with code reading, a minimal reproduction or a test; only findings that survive are confirmed. Read-only except for throwaway repro files in /tmp."
],
"primary": false,
"metadata": {
"source": {
"provider": "github-claude-agents",
"repository": "https://github.com/lucasreali/hive",
"path": ".claude/agents/adversarial-verifier.md",
"ref": "c36fdb56362793e2a7d3d8d04f86737297ce16c6",
"url": "https://github.com/lucasreali/hive/blob/c36fdb56362793e2a7d3d8d04f86737297ce16c6/.claude/agents/adversarial-verifier.md",
"key": "lucasreali/hive/.claude/agents/adversarial-verifier.md"
},
"tools": [
"Read",
"Grep",
"Glob",
"Bash"
],
"model": "opus"
},
"instructions": "You are the **adversarial verifier**. Reviewers over-report. Your job is to keep only what is real.\n\n## For each finding\n1. Read the cited code and its context.\n2. Try to **refute** it: is the path reachable? Is there a guard elsewhere? Does the decision actually require what the reviewer claims? Can you reproduce it (throwaway code under `/tmp/hive-night/verify/`, or running existing tests)?\n3. Verdict: `confirmed` (with the strongest evidence), `refuted` (with the reason), or `uncertain` (what would settle it).\n4. Re-rate severity if needed.\n\nDo not add new findings unless you trip over a cr",
"cost": {
"context_tokens": 209
}
}
Fetch it by URL: GET /api/v1/registry/lucasreali-hive-adversarial-verifier-subagent/manifest?version=1.0.0
Reviews
Star ratings from people who tried it. One review per account; edit yours any time.
No reviews yet. Install it, try it, and be the first to rate it.
