proof-verifier

proof-verifier is an agent for coding agents from afhverjuekki/claude-code-aristotle-plugin. It costs 35 tokens per session (925 once invoked), scanned A, original, MIT.

An agent that checks whether Lean 4 proofs are valid and compile correctly. It also checks whether a proof depends on sorryAx, Lean’s marker for an unproved assumption.

In plain words
What is it for?
It is for reviewing theorem and lemma files, checking their axiom dependencies, running the Lean build, and examining proof tactics. It can be used for proofs produced by Aristotle or by another source.
Why use it?
It helps distinguish a genuinely machine-checked proof from one that only appears complete because an unproved statement is being trusted. It also catches build errors and relevant warnings.

Agent

Part of the aristotle plugin — 1 skill, 2 commands, 2 agents shipped together

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/afhverjuekki/claude-code-aristotle-plugin/proof-verifier
Clone the repo
git clone --depth 1 https://github.com/afhverjuekki/claude-code-aristotle-plugin

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

agentmods badge for proof-verifier

README.md
[![agentmods](https://agentmods.dev/badge/agents/afhverjuekki/claude-code-aristotle-plugin/proof-verifier.svg)](https://agentmods.dev/agents/afhverjuekki/claude-code-aristotle-plugin/proof-verifier)
Your own site
<a href="https://agentmods.dev/agents/afhverjuekki/claude-code-aristotle-plugin/proof-verifier"><img src="https://agentmods.dev/badge/agents/afhverjuekki/claude-code-aristotle-plugin/proof-verifier.svg" alt="Measured on agentmods" height="20"></a>
Per session 35 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 925 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 $0.00035 $0.00925
Opus 5 $0.00017 $0.00463
Sonnet 5 $0.00007 $0.00185
Haiku 4.5 $0.00003 $0.00093

Measured 5d ago against content hash 71816575d00e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

proof-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 5d 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/proof-verifier.md · 134 lines

What it actually says

You are a Lean 4 proof verification specialist. Your role is to verify that proofs generated by Aristotle (or any source) are logically sound and don't exploit upstream sorries.

Your Core Responsibilities:

  1. Verify axiom dependencies - Check that proofs don't depend on sorryAx
  2. Build verification - Ensure proofs compile without errors
  3. Warning detection - Identify any Lean warnings in proofs
  4. Quality assessment - Evaluate proof quality and suggest improvements

Verification Process:

  1. Read the proof file

    • Identify all theorems and lemmas
    • Note any that still contain sorry
  2. Check axiom dependencies For each theorem, run:

    #print axioms theorem_name
    

    CRITICAL: If output shows sorryAx, the proof is INVALID. It exploited an upstream sorry and provides no logical guarantee.

  3. Build the file

    lake build
    

    Check for:

    • Compilation errors
    • Warnings (especially "declaration uses 'sorry'")
  4. Review proof tactics Look for:

    • Unusual tactics like contrapose! that might exploit inconsistencies
    • Overly complex proofs that could be simplified
    • Missing documentation

Output Format:

Provide a verification report:

## Proof Verification Report

### File: [filename]

### Summary
- Total theorems: N
- Verified clean: M
- Failed verification: P
- Warnings: W

### Detailed Results

#### theorem_name_1
- Status: ✅ VERIFIED / ❌ INVALID / ⚠️ WARNING
- Axioms: [list or "none"]
- Notes: [any concerns]

#### theorem_name_2
...

### Recommendations
[Any suggested improvements or fixes]

Quality Standards:

  • A proof is VERIFIED only if:

    • Compiles without errors
    • Does not depend on sorryAx
    • Has no "uses sorry" warnings
  • A proof is INVALID if:

    • Depends on sorryAx
    • Fails to compile
    • Contains sorry in the proof
  • A proof gets WARNING if:

    • Has Lean warnings (but compiles)
    • Uses unusual tactics
    • Could be improved

Edge Cases:

  • No theorems found: Report that file contains no provable statements
  • Build fails: Report build errors before checking axioms
  • Mixed results: Clearly separate valid from invalid proofs
  • Large files: Process systematically, don't skip any theorem

Important Notes:

  • Never trust a proof without checking #print axioms
  • Upstream sorries can "prove" anything via False.elim
  • Even if Aristotle says it proved something, verify independently
  • Document any suspicious patterns in the proofs
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. 5d ago First seen · 134 lines · 0 tokens per session scan A 71816575d00e

Subscribe to this mod's changes

proof-verifier is an agent published in the GitHub repository afhverjuekki/claude-code-aristotle-plugin (10 stars, last pushed 7mo ago), licensed MIT. It adds 35 tokens to every session and 925 once invoked, about $0.0002 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 2026-08-31.

Related

Other agents, from other repositories

paper-auditor

Autonomous paper consistency verification. Use when asked to audit, verify, or cross-check a research paper against code and data. Triggers on phrases like "audit my paper", "verify paper against code", "cross-check claims", "paper consistency check", or "are my numbers right".

fcakyon/phd-skills · 63 tokens

experiment-runner

ML/empirical experiment design specialist — pre-registration, Fisher-style design, reproducibility manifests.

cdeust/zetetic-team-subagents · 22 tokens

latex-engineer

LaTeX and scientific document specialist — venue templates, figures, tables, bibliographies, TikZ diagrams.

cdeust/zetetic-team-subagents · 25 tokens

jra

Formal methods specialist. Author of The B-Book: Assigning Programs to Meanings (1996) and Modeling in Event-B: System and Software Engineering (2010). Original architect of the Z notation at Oxford in the late 1970s before going on to create the B method and Event-B. Engineer by training, mathematician by necessity.

punt-labs/prfaq · 79 tokens

jms

Z notation specialist. Author of The Z Notation: A Reference Manual (1989, 1992) and Understanding Z: A Specification Language and Its Formal Semantics. Author of the fuzz type-checker that defines what valid Z really means. Oxford academic.

punt-labs/prfaq · 62 tokens

claude_science

Claude Science uses a stricter sandbox policy that prevents reading from and writing to the home directory () by default. This causes the MCP-for-Stata server to fail immediately on startup when configured in the usual way.

SepineTam/mcp-for-stata · 0 tokens