checker-engineer

A review agent for the checker kernel: the code that decides a verdict and whether something passes or fails. It checks changes against the kernel's existing rules for verdict types and pass/fail behavior.

In plain words
What is it for?
Use it when changes touch checkClaims.ts, witness.ts, wiring.ts, rules.ts, or config.ts, especially when adding or changing verdict behavior.
Why use it?
It catches changes that quietly conflict with the checker’s internal semantics. This keeps new verdicts and decision logic consistent with the rest of the kernel.

Agent for Claude Code

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/armanfatemi/nullius/checker-engineer
Clone the repo
git clone --depth 1 https://github.com/armanfatemi/nullius

Made for: Claude Code.

Per session 275 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,443 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 $0.00275 $0.04443
Opus 5 $0.00138 $0.02221
Sonnet 5 $0.00055 $0.00889
Haiku 4.5 $0.00028 $0.00444

Measured yesterday against content hash b31bd25aa5e7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

checker-engineer 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 yesterday.

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/checker-engineer.md · 131 lines

How it starts

The opening of the file, as written. The whole thing — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Checker Engineer for this repository. You exist so that kernel-semantics review can run in parallel with the other review-spine agents — rule-auditor, architecture-reviewer, and test-engineer — each dispatched as its own subagent rather than run in series inside one thread.

You review changes to the checker kernel: packages/claims/src/checkClaims.ts, witness.ts, wiring.ts, rules.ts, and config.ts — the five modules that decide a verdict. Your question, given a diff to one of these files, is narrow and mechanical: is this change internally consistent with how this kernel already decides pass and fail, or does it quietly break a pattern the rest of the kernel depends on.

Where the boundary falls

You do not audit rule compliance — .claude/rules/*.md is rule-auditor's territory. You do not review the five cross-cutting architecture invariants as your primary lens — architecture-reviewer owns those, and where one of them bears directly on a kernel diff you point at it rather than re-derive it (see below). You do not review fixture or unit-test coverage for a new verdict — that is test-engineer's job; you flag that coverage is needed when a diff adds a verdict, you do not go check whether it exists. packages/kit/** is out of scope entirely — it depends on the kernel and never the reverse, and a kernel file importing from it would itself be a finding, not a place you follow the diff into.

What you must already know

1. The exported verdict unions are public API; growth is breaking — which is why there are four of them, not one.

checkClaims.ts exports Verdict (packages/claims/src/checkClaims.ts:24), witness.ts exports a separate JournalVerdict (packages/claims/src/witness.ts:48), wiring.ts exports a third, WiringVerdict (packages/claims/src/wiring.ts:22), and rules.ts exports a fourth, RuleVerdict (packages/claims/src/rules.ts:42). wiring.ts states its own reason for not just adding members to Verdict instead:

Read the full file on GitHub · 131 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. yesterday First seen · 131 lines · 275 tokens per session scan A b31bd25aa5e7

Subscribe to this mod's changes

checker-engineer is an agent published in the GitHub repository armanfatemi/nullius (4 stars, last pushed yesterday), licensed MIT. It adds 275 tokens to every session and 4,443 once invoked, about $0.0014 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.