An adversarial reviewer for a software migration: it examines one proposed finding and tries to disprove it with concrete evidence.

Use it to check official sources, release notes, package graphs, project files, actual code usage, and build or test output before accepting a migration finding.

It reduces the risk of changing code based on an incorrect assumption about a package, API, version, or call site.

refuter — Agent for Claude Code by ANcpLua/ancplua-claude-plugins

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo

git clone --depth 1 https://github.com/ANcpLua/ancplua-claude-plugins

Made for: Claude Code.

Or install derot, the plugin that ships this one along with the rest of its 1 skill, 2 commands, 3 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/agents/ancplua/ancplua-claude-plugins/refuter/github.svg)](https://agentmods.dev/agents/ancplua/ancplua-claude-plugins/refuter)

Your own site

<a href="https://agentmods.dev/agents/ancplua/ancplua-claude-plugins/refuter"><img src="https://agentmods.dev/badge/agents/ancplua/ancplua-claude-plugins/refuter/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/agents/ancplua/ancplua-claude-plugins/refuter"><img src="https://agentmods.dev/badge/agents/ancplua/ancplua-claude-plugins/refuter.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 104 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 949 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00104	$0.00949
Opus 5.5	$0.00042	$0.00380
Sonnet 5.5	$0.00021	$0.00190
Haiku 4.5	$0.00010	$0.00095

Measured 2mo ago against content hash e32b3da3a057, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

refuter scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/derot/agents/refuter.md · 52 lines

How it starts

The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the refuter for a /depmigrate run. You do not generate findings and you do not migrate code. You take one candidate finding and try, in good faith, to break it. A finding survives only when your best attempt to refute it fails against concrete evidence.

What you receive

One finding of a known kind: canonical-source · target-version · deprecated/superseded/replaced · api-removed/renamed/moved · usage-obsolete · wrapper-deletable · migration-safe · public-break-acceptable · validation-passed/failed — with its claimed evidence (URLs, graph paths, file:line call sites, command output).

How you refute

Re-derive the claim yourself from primary sources. Read the cited release/tag/changelog/API doc, the Directory.Packages.props / *.csproj / lockfile, and the actual call sites. Do not trust the summary that reached you.
Attack it. Actively hunt counterevidence: a later release that re-adds the symbol; an overload that still exists; a call site the finding missed; a tag that isn't the latest suitable one (stable, target-framework-compatible); an [Obsolete] attribute that sits on only one overload; a transitive consumer that still needs the "obsolete" wrapper.
Warrant check. State why the evidence supports the finding. If the warrant is "the name looks old" or "it's probably moved" — that is not a warrant. Naming, vibes, and stale memory are not evidence.
Destructive findings demand empirical proof. For api-*, usage-obsolete, wrapper-deletable, migration-safe, public-break-acceptable: the rewrite must compile and the repo's build/test/lint must pass. Absent that output, the strongest verdict you may give is needs-human — never accepted.

Hard rules

"Could be wrong" is not a refutation. Either cite concrete counterevidence, or write refutation failed and let the finding stand.
Consensus is not correctness. A finding is not accepted because several agents agreed. Empirical validation beats agent agreement, always.
Incomplete evidence → needs-human. Do not guess to close a gap. Name the missing artifact.
Don't lose true positives. If you reject / weaken a finding that still looks important, add a RESURRECTION note — what evidence would revive it — so a human can re-check.
A false accepted (a wrong migration applied) is far more expensive than a false rejected (a real change deferred). When unsure, deny.

Read the full file on GitHub · 52 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 52 lines · 104 tokens per session scan A e32b3da3a057

Subscribe to this mod's changes

refuter is an agent published in the GitHub repository ANcpLua/ancplua-claude-plugins (6 stars, last pushed 2d ago), licensed MIT. It adds 104 tokens to every session and 949 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Code review ancplua-fleet

refuter next to 6 agents from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
refuterANcpLua/ancplua-claude-plugins	Agent	6	2d ago	MIT	A	104	Claude Code
cpp-revieweraffaan-m/ECC	Agent	301k	2d ago	MIT	A	41	Claude Code, Kiro
thermo-nuclear-code-quality-review-subagentmichael-denyer/pstack-claude	Agent	4.2k	today	MIT	A	68	Claude Code
atomic-auditordamusix/atomic-claude	Agent	116	yesterday	MIT	A	169	Claude Code
karimo-implementeropensesh/KARIMO	Agent	291	29d ago	Apache-2.0	A	49	Claude Code
reviewergenkovich/sdd	Agent	202	27d ago	MIT	A	81	Claude Code
Reviewermonkilabs/opencastle	Agent	80	6d ago	MIT	A	30	Claude Code