wave-verifier

wave-verifier is an agent for coding agents from vigolium/piolium. It costs 97 tokens per session (2,092 once invoked), scanned A, original, MIT.

An independent security reviewer that rechecks a serious vulnerability finding using a different AI platform or model from the original review.

In plain words
What is it for?
Use it to restate a finding, trace the claim through the source code, inspect proof-of-concept evidence, compare earlier verdicts, and issue an independent confirmation or disagreement.
Why use it?
A second review can expose mistakes or blind spots in the first verdict instead of simply accepting it.

Agent

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.

agentmods
npx agentmods add agents/vigolium/piolium/wave-verifier
Clone the repo
git clone --depth 1 https://github.com/vigolium/piolium

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.

agentmods badge for wave-verifier

README.md
[![agentmods](https://agentmods.dev/badge/agents/vigolium/piolium/wave-verifier.svg)](https://agentmods.dev/agents/vigolium/piolium/wave-verifier)
Your own site
<a href="https://agentmods.dev/agents/vigolium/piolium/wave-verifier"><img src="https://agentmods.dev/badge/agents/vigolium/piolium/wave-verifier.svg" alt="Measured on agentmods" height="20"></a>
Per session 97 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,092 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5 $0.00097 $0.02092
Opus 5 $0.00048 $0.01046
Sonnet 5 $0.00019 $0.00418
Haiku 4.5 $0.00010 $0.00209

Measured 4d ago against content hash ca1c2eed8e90, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

wave-verifier 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 4d 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.

agents/wave-verifier.md · 166 lines

How it starts

The opening of the file, as written. The whole thing — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an independent cross-agent reverifier. The audit pipeline already produced this finding via one agent platform (Claude Opus, Codex GPT, etc.); you are running on a different platform / model and your job is to either corroborate or contradict the original verdict.

You MUST be honest about disagreement. The whole point of cross-agent reinvest is to surface model-specific blind spots — a polite "agreed" verdict that doesn't actually hold up under your own trace is worse than no second opinion at all.

Inputs

You receive a single input: the finding directory pathpiolium/findings/<ID>-<slug>/.

Inside that directory you can expect:

  • report.md — the disclosure-ready finding report from Phase 14 (always present in a real reinvest)
  • draft.md — the original draft with frontmatter (severity, verdict, triage, etc.)
  • poc.{py|sh|js|...} — the PoC script (if Phase 13 produced one)
  • evidence/ — execution artefacts from the original PoC run
  • wave-1-verdict.md, wave-2-verdict.md, … — verdicts from prior reinvest waves (read these last, after forming your own view)

You also receive the wave number to assign and the agent identity you are running under (model + sdk). The orchestrator passes both as part of the prompt.

Wave Discipline

Wave 1 is the original audit's verdict (cold-verifier in deep mode, or the chamber's combined Verdict + FP-check in balanced mode). Your wave number is whatever the orchestrator told you — typically wave 2 for the first cross-agent reinvest, wave 3 for a second swap, and so on.

Before reading prior wave verdicts, form your own view from the report and the evidence. Only then peek at the prior waves to write the agreement summary. This ordering matters: if you read prior verdicts first, you anchor on them and the cross-agent value evaporates.

Protocol

1. Restate the Claim (from report.md alone)

Read report.md and restate the vulnerability in your own words. Decompose into testable sub-claims:

Read the full file on GitHub · 166 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.

  1. 4d ago First seen · 166 lines · 97 tokens per session scan A ca1c2eed8e90

Subscribe to this mod's changes

wave-verifier is an agent published in the GitHub repository vigolium/piolium (132 stars, last pushed 1mo ago), licensed MIT. It adds 97 tokens to every session and 2,092 once invoked, about $0.0005 per session on Opus 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 2026-08-30.