claude-code-build-auditor

claude-code-build-auditor is a skill for Claude Code, Codex from jpeslar1/john-peslar-ai-skills. It costs 149 tokens per session (1,732 once invoked), scanned A, original, MIT.

An adversarial review process for software built with AI coding tools, including its instructions, architecture, plans, and code changes. It looks for unnecessary complexity, weak guidance, and missing verification.

In plain words
What is it for?
Assess architecture before implementation, prepare major agent tasks, investigate implementation drift, and improve instructions, tools, tests, or permissions when repeated problems appear.
Why use it?
It helps catch problems that can make an AI coding agent produce bloated, incorrect, or poorly checked work. The review is based on observed failures and the actual codebase.

Skill for Claude CodeCodex

Written for Claude Code and Codex: shipped in a Claude Code plugin, but also agents/openai.yaml present. Also seen: mentions Claude Code; mentions Codex.

Part of the ai-automation plugin — 2 skills shipped together

Good fit Assess architecture before implementation, prepare major agent tasks, investigate implementation drift, and improve instructions, tools, tests, or permissions when repeated problems appear.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/jpeslar1/john-peslar-ai-skills/claude-code-build-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.

Any agent
npx skills add jpeslar1/john-peslar-ai-skills --skill claude-code-build-auditor
Clone the repo
git clone --depth 1 https://github.com/jpeslar1/john-peslar-ai-skills

Made for: Claude Code, Codex.

Or install ai-automation, the plugin that ships this one along with the rest of its 2 skills.

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 claude-code-build-auditor

README.md
[![agentmods](https://agentmods.dev/badge/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor/github.svg)](https://agentmods.dev/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor)
Your own site
<a href="https://agentmods.dev/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor"><img src="https://agentmods.dev/badge/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for claude-code-build-auditor

Your own site · 80×15
<a href="https://agentmods.dev/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor"><img src="https://agentmods.dev/badge/skills/jpeslar1/john-peslar-ai-skills/claude-code-build-auditor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 149 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,732 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 $0.00149 $0.01732
Opus 5 $0.00075 $0.00866
Sonnet 5 $0.00030 $0.00346
Haiku 4.5 $0.00015 $0.00173

Measured 12d ago against content hash c633604dd460, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

claude-code-build-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 12d 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-code-build-auditor/SKILL.md · 170 lines

How it starts

The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Claude Code Build Auditor

Use this skill to audit an AI-assisted build like a Claude Code product engineer: delete unnecessary scaffolding, observe the model against the real codebase, add back only proven instructions, and require verification loops before trusting output.

Read references/source-map.md when the user asks where the framework came from, wants the Boris Cherny talk summarized, or wants the skill updated from a new Claude Code talk.

Core Principle

Modern coding agents often fail because the product, prompt, or workflow gets in the model's way. Do not solve every failure by adding more instructions. Start by removing bloat, run the model on the real task, observe repeated failure patterns, then add the smallest durable context, test, tool, permission, or skill that fixes the repeated failure.

When To Use This Skill

Use it at five points in a build:

  1. Before architecture - Decide whether the architecture is simple enough for an agent and human maintainer to reason about.
  2. Before a big agent run - Give the agent the task, guardrails, exit criteria, and verification path without overspecifying every step.
  3. During implementation drift - Stop AI bloat, duplicate abstractions, unnecessary files, speculative framework changes, and instruction sprawl.
  4. After a model or tool upgrade - Re-test old prompts, skills, hooks, and MCP context. Newer models may not need old scaffolding.
  5. Before shipping - Verify the work through tests, product behavior, logs, screenshots, readbacks, or a focused code review.

Audit Lenses

Apply these in order.

  1. Delete-first ablation

    • What instructions, abstractions, hooks, helper files, wrappers, or process steps can be removed?
    • What happens if the agent runs with less instruction?
    • Is this line here because the current model needs it, or because an older model once needed it?
  2. Repeated-failure proof

    • Do not add permanent instructions after one weird failure.
    • Add context only when the model repeatedly stumbles on the same project-specific issue.
    • Prefer a small skill, test, comment, or MCP data source over a giant prompt.

Read the full file on GitHub · 170 lines

Files

What ships with it

2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. 12d ago First seen · 170 lines · 149 tokens per session scan A c633604dd460

Subscribe to this mod's changes

claude-code-build-auditor is a skill published in the GitHub repository jpeslar1/john-peslar-ai-skills (6 stars, last pushed 3d ago), licensed MIT. It adds 149 tokens to every session and 1,732 once invoked, about $0.0007 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.