lattice-auditor

lattice-auditor is an agent for Claude Code from plausibleventures/lattice. It costs 42 tokens per session (545 once invoked), scanned A, original, MIT.

An adversarial reviewer for Lattice packages, which are reusable parts of a TypeScript game kit. It checks completed work and reports real defects, fixing only what its task allows.

In plain words
What is it for?
Use it to test package behavior with unusual inputs, verify deterministic output, check stated rules, inspect performance-sensitive code, and compare documentation with the actual API.
Why use it?
It looks beyond ordinary tests for edge cases, non-repeatable results, broken rules, unnecessary per-frame work, and examples that fail for a new user.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; mentions AGENTS.md.

Part of the lattice plugin — 12 skills, 1 command, 6 agents shipped together

Good fit Use it to test package behavior with unusual inputs, verify deterministic output, check stated rules, inspect performance-sensitive code, and compare documentation with the actual API.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/plausibleventures/lattice/lattice-auditor
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/plausibleventures/lattice

Made for: Claude Code.

Or install lattice, the plugin that ships this one along with the rest of its 12 skills, 1 command, 6 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 lattice-auditor

README.md
[![agentmods](https://agentmods.dev/badge/agents/plausibleventures/lattice/lattice-auditor.svg)](https://agentmods.dev/agents/plausibleventures/lattice/lattice-auditor)
Your own site
<a href="https://agentmods.dev/agents/plausibleventures/lattice/lattice-auditor"><img src="https://agentmods.dev/badge/agents/plausibleventures/lattice/lattice-auditor.svg" alt="Measured on agentmods" height="20"></a>
Per session 42 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 545 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00042 $0.00545
Opus 5 $0.00021 $0.00272
Sonnet 5 $0.00008 $0.00109
Haiku 4.5 $0.00004 $0.00055

Measured 8d ago against content hash fc968ec442f0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

lattice-auditor 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 8d 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.

.claude/agents/lattice-auditor.md · 39 lines

What it actually says

You audit Lattice. Your job is to find what is wrong, and you are measured on real findings, not on volume. Read AGENTS.md; every rule in it is something you check.

Audit in this order, hardest first:

  1. Correctness under adversarial input. Zero, one, negative, NaN, Infinity, empty array, single element, exact boundary, huge magnitude, and the value that makes a divisor zero. Read the tests and ask what they cannot catch, then write the case that proves it.
  2. Determinism. Same seed, same inputs, same bytes — across two runs in one process and across a fresh process. Grep for the banned non-deterministic sources and for anything that iterates a Set or object whose order depends on insertion history.
  3. The stated invariants. .lattice/kit.json lists them per package. Each one is a claim. Try to falsify it with a test.
  4. Allocation on the hot path. Anything per-frame or per-entity that returns an object, builds a closure, or .map()s an array is a finding.
  5. The API as a stranger meets it. Take the README example, follow it exactly, and see whether it works. Then try the obvious wrong thing and check the error message names the caller's mistake.
  6. Docs against behaviour. A comment that is no longer true is a defect of the same severity as a wrong branch, because the next agent will trust it.

Rules of evidence: a finding is a file, a line, an input, and the wrong output it produces. "This could be a problem" is not a finding. Verify before you report — try to refute your own claim first, and drop it if it does not survive.

If your task authorises fixes, fix only what you found, keep each fix minimal, and leave npm run verify green. If it does not, report and change nothing.

Your final message is a ranked list, worst first, each with the reproduction. If you found nothing real, say that plainly — a clean audit reported honestly is worth more than six invented nits.

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. 8d ago First seen · 39 lines · 42 tokens per session scan A fc968ec442f0

Subscribe to this mod's changes

lattice-auditor is an agent published in the GitHub repository plausibleventures/lattice (37 stars, last pushed 15d ago), licensed MIT. It adds 42 tokens to every session and 545 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 2026-08-30.

Related

Other agents, from other repositories

godot-code-reviewer

Use this agent when the user wants their Godot GDScript or C# code reviewed for best practices, anti-patterns, performance issues, or Godot-specific pitfalls. Also use when completing a major feature and wanting a quality check. Examples: Context: User wants a code review. user: "Review my player controller for Godot…

jame581/GodotPrompter · 187 tokens

godot-reviewer

Reviews Godot architecture, scene structure, resource usage, and engine-specific risks.

MRCalderon3D/everything-game-dev-code · 21 tokens

reviewer

Post-implementation code reviewer for Godot game projects. Reads implemented code, decides which domain-specific reviewer skills apply, runs their checklists, and reports issues found. MUST NOT modify project files.

RandallLiuXin/GodotMaker · 42 tokens

gamemaker-gml-specialist

The GML Specialist is the hands-on GML coding authority. They write, review, and refactor GML code with deep knowledge of language features, patterns, style, memory management, and the full GML API surface.

TraftG/opencode-game-studio · 46 tokens

file-verifier

External semantic verifier for a single Godot file (.gd / .tscn / .tres / .gdshader). Reads the file fresh — no context from the writer — checks against Godot 4.x API via godot-docs MCP, project conventions, and the skill the writer claimed to follow. Returns findings only; does not rewrite. Invoked after a batch of…

Simone-Tarantino/godot-superpowers · 110 tokens

code-reviewer

Review GDScript and scene files against Godot 4.x best practices — type hints, signals, composition, performance hotspots, and Godot 3.x leftovers. Use after any non-trivial change or before committing.

Simone-Tarantino/godot-superpowers · 49 tokens