ai-eng-warden

ai-eng-warden is an agent for Claude Code from sliamh11/Deus. It costs 0 tokens per session (1,709 once invoked), scanned A, original, MIT.

A strict reviewer for changes involving artificial-intelligence systems, such as language-model prompts, context handling, agent design, retrieval systems, or token limits. It checks the quality, efficiency, architecture, and security of those AI-specific parts.

In plain words
What is it for?
It is for reviewing diffs, meaning proposed code changes, that modify prompt templates, agent instructions, retrieval or RAG code, model parameters, gate specifications, or token-budget logic.
Why use it?
It helps catch problems in prompts, context use, model settings, and AI security before the changes are accepted. In strict mode, it can require revisions before a commit proceeds.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter.

Good fit It is for reviewing diffs, meaning proposed code changes, that modify prompt templates, agent instructions, retrieval or RAG code, model parameters, gate specifications, or token-budget logic.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/sliamh11/deus/ai-eng-warden
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/sliamh11/Deus

Made for: Claude Code.

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 ai-eng-warden

README.md
[![agentmods](https://agentmods.dev/badge/agents/sliamh11/deus/ai-eng-warden.svg)](https://agentmods.dev/agents/sliamh11/deus/ai-eng-warden)
Your own site
<a href="https://agentmods.dev/agents/sliamh11/deus/ai-eng-warden"><img src="https://agentmods.dev/badge/agents/sliamh11/deus/ai-eng-warden.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,709 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00000 $0.01709
Opus 5 $0.00000 $0.00855
Sonnet 5 $0.00000 $0.00342
Haiku 4.5 $0.00000 $0.00171

Measured 8d ago against content hash a96237957fb6, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

ai-eng-warden scanned grade A with 1 finding 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.

Runs shell commandslowCapability

Expected in a hook, worth knowing in a rule or an instructions file.

subprocess.run([sys.executable, 'evolution/cli.py', 'dismiss_warden_finding', payload])
.claude/agents/ai-eng-warden.md · 98 lines

How it starts

The opening of the file, as written. The whole thing — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the ai-eng-warden — a specialized AI Engineering reviewer for code that touches LLM interactions. You review like a senior AI engineer: prompt quality, context management, architecture decisions, token efficiency, and AI-specific security. You do NOT review general code quality (that's code-reviewer's job). You focus exclusively on the AI engineering dimensions.

At invocation, read these (be surgical)

  1. Standards~/deus/.claude/wardens/standards.md. Sets the quality floor for all wardens.
  2. Rules file (primary)~/deus/.claude/wardens/ai-engineering-rules.md. Read the routing tier first (everything above ## Remediation Details). Apply every rule whose Applies when matches the diff. For rules that fire, read the matching detail block below ## Remediation Details for Remediation. Scope memo: If .claude/.warden-memo.md exists, read it FIRST before steps 3-6. It was written by code-reviewer and contains pre-discovered scope context.
  3. The diff or file list — determine mode from the invocation prompt:
    • Audit mode: if the prompt contains AUDIT MODE: followed by a file list, treat each listed file as the review target. Skip git diff entirely. Jump to step 4 using those files. Output line budget is ≤120 lines in audit mode. Also skim ~/deus/docs/decisions/INDEX.md for ADR orientation — use it to avoid re-litigating settled architecture, not to enforce compliance.
    • Diff mode (default): resolve diff from prompt or cwd:
      • If the prompt cites a worktree path, use it: git -C <worktree> diff and git -C <worktree> diff --cached.
      • Otherwise: git diff and git diff --cached.
      • If BOTH empty → "no changes to review" and stop.
  4. Full prompt templates — unlike code-reviewer, you MUST read the full files (not just diff hunks) when the diff touches prompt construction, gate specs, or role specs. Context positioning and overall prompt structure matter.
  5. Gate specs~/deus/.claude/agents/wardens/*.md — read when diff touches gate evaluation code.
  6. Role specs~/deus/.claude/agents/*.md — read when diff touches agent dispatch code.

Read the full file on GitHub · 98 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. 8d ago First seen · 98 lines · 0 tokens per session scan A a96237957fb6

Subscribe to this mod's changes

ai-eng-warden is an agent published in the GitHub repository sliamh11/Deus (51 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,709 tokens. A static security scan graded it A with 1 finding (runs shell commands). 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

evaluator-data

Commander Data evaluator for logical analysis and edge cases. Focus: logical consistency, edge cases, unintended consequences. Domain evaluator - invoked for technical and analytical tasks.

datacore-one/datacore · 35 tokens

evaluator-cto

Evaluates from CTO/technical perspective. Focus: technical accuracy, best practices, scalability. Core evaluator - always runs for every task.

datacore-one/datacore · 32 tokens

evaluator-dijkstra

Dijkstra evaluator for code and algorithmic thinking. Focus: correctness, elegance, simplicity. Domain evaluator - invoked for :AI:code: tasks.

datacore-one/datacore · 35 tokens

prompt-reviewer

LLM prompt-engineering expert for the review-panel skill. Spawned when the diff touches LLM/API prompts, prompt templates, or inline model instructions in application code (system/user prompts, few-shot templates, prompt-string builders). Reviews prompting quality, output contracts, context economy, injection surface…

kbichave/skills · 92 tokens

ai-ethics-reviewer

AI / ML ethics + responsible-AI specialist — bias, fairness, model selection, dataset provenance, automated-decision systems, AI safety, prompt injection defense, model cards. Use PROACTIVELY on any ML / AI / LLM / generative-AI work. Owns Council Division 15.

Nmor/the-claude-council · 69 tokens

mle-reviewer

Production machine-learning engineering reviewer for data contracts, feature pipelines, training reproducibility, offline/online evaluation, model serving, monitoring, and rollback. Use when ML, MLOps, model training, inference, feature store, or evaluation code changes.

Fmarzochi/EGC · 55 tokens