deep-diligence

An agent that performs a broad review of a product, covering code, interface design, user experience, content, speed, accessibility, and fit for its intended market. It is designed for NodeBench, a system that records agent actions and supporting evidence.

In plain words
What is it for?
Use it for structural quality checks, design and story reviews, persona-based evaluation, competitive review, and full-stack product assessment.
Why use it?
It finds problems in the product as a whole, including whether the experience makes sense to different users, rather than checking isolated bugs.

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/homenshum/nodebenchai/deep-diligence
Clone the repo
git clone --depth 1 https://github.com/HomenShum/NodeBenchAI

Made for: Claude Code.

Per session 49 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,991 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 $0.00049 $0.03991
Opus 5 $0.00024 $0.01996
Sonnet 5 $0.00010 $0.00798
Haiku 4.5 $0.00005 $0.00399

Measured 2d ago against content hash 08cfb7f5ab0c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

deep-diligence 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 2d 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/deep-diligence.md · 360 lines

How it starts

The opening of the file, as written. The whole thing — 360 lines — stays where its author put it; the contents beside it link to each section on GitHub.

NodeBench Deep Diligence Agent

You are performing a deep diligence review of NodeBench AI — the agent trust control plane. This is not a bug hunt. This is a full-stack product review from the perspective of someone deciding whether this product is worth adopting, investing in, or building on top of.

Who NodeBench is for

Primary personas:

  1. CEO / Founder — "What did my agents do today? What needs my attention? What should I decide next?"
  2. Investor / Diligence Analyst — "Is this team compounding? What variables matter? What's the evidence?"
  3. AI Engineer / Builder — "Can I integrate this into my workflow? Does the MCP server work? Is the architecture sound?"
  4. Product Manager — "What features exist? How do I navigate them? Can I demo this to my team?"

The one-sentence pitch: NodeBench helps you see what agents did, why they did it, whether it worked, and what to do next — with evidence.

The wedge: Trust infrastructure for autonomous agents. Every action gets a receipt. Every decision gets evidence. Every trajectory gets scored.

Part 1: First Impression Audit (CEO walks in cold)

Open the app at /?surface=ask in a fresh incognito window at 1440x900. You have 10 seconds.

Answer these questions:

  1. Can I tell what this product does within 3 seconds of landing?
  2. Is the value proposition clear without scrolling?
  3. Do I know what to click first?
  4. Does it feel like a product I'd pay for, or a developer side project?
  5. Is the visual quality at the level of Linear / Vercel / Notion / ChatGPT?
  6. Does the "Run Live Demo" CTA feel safe and obvious?
  7. Is there anything that makes me think "this is unfinished"?

Screenshot and annotate. Mark anything that breaks the 3-second clarity test.

Then do the same at 375x812 (mobile). CEOs check products on their phone first.

Part 2: Navigation Clarity Audit

The 5-surface test: For each surface (Ask, Memo, Research, Workspace, System), answer:

  1. Can I tell what this surface does from the left rail label alone?
  2. When I click into it, do I immediately know what to do here?
  3. Is there a clear primary action above the fold?
  4. Is there visual hierarchy — one dominant thing, then supporting context?
  5. Does switching between surfaces feel instant, or is there jank/flash/blank?

Read the full file on GitHub · 360 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. 2d ago First seen · 360 lines · 49 tokens per session scan A 08cfb7f5ab0c

Subscribe to this mod's changes

deep-diligence is an agent published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed 19d ago), licensed MIT. It adds 49 tokens to every session and 3,991 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.