dogfood-loop

A testing workflow that uses a deployed NodeBench app as a real user would, then scores its screens and tools and records problems.

In plain words
What is it for?
It is for reviewing the Decision Workbench, postmortem, agent telemetry, and MCP tools, with screenshots, scores, and filed findings.
Why use it?
It helps reveal unclear screens, missing information, slow loading, and other issues through structured hands-on checks.

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/dogfood-loop
Clone the repo
git clone --depth 1 https://github.com/HomenShum/NodeBenchAI

Made for: Claude Code.

Per session 25 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 778 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 $0.00025 $0.00778
Opus 5 $0.00013 $0.00389
Sonnet 5 $0.00005 $0.00156
Haiku 4.5 $0.00003 $0.00078

Measured 2d ago against content hash 08354f1392d1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

dogfood-loop 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/dogfood-loop.md · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the NodeBench self-dogfood agent. Your job is to USE the product as a real user would, score the experience, and file findings.

Dogfood Protocol

Phase 1: Decision Workbench Test (3 min)

  1. Navigate to the deployed app (use preview or Chrome MCP)
  2. Go to /deep-sim (Decision Workbench)
  3. Screenshot the page
  4. Evaluate: Does the memo answer the question above the fold? Are variables visible? Are scenarios clear? Is confidence shown?
  5. Score 1-5 on: clarity, evidence density, actionability, visual quality, load time

Phase 2: Postmortem Test (2 min)

  1. Navigate to /postmortem
  2. Screenshot
  3. Evaluate: Is the prediction-vs-reality comparison clear? Are scoring dimensions visible? Is "what we learned" useful?
  4. Score 1-5 on: comparison clarity, scorecard readability, actionable learning, visual quality

Phase 3: Agent Telemetry Test (2 min)

  1. Navigate to /agent-telemetry
  2. Screenshot
  3. Evaluate: Can I see total actions, tools used, cost, latency at a glance? Is the table sortable? Are errors highlighted?
  4. Score 1-5 on: data density, scanability, cost visibility, error surfacing

Phase 4: MCP Tool Quality Test (3 min)

  1. Run extract_variables via MCP for entity "product/nodebench-ai"
  2. Run score_compounding for the same entity
  3. Evaluate: Did the tools return structured data? Was confidence included? Was "whatWouldChangeMyMind" present?
  4. Score 1-5 on: response structure, provenance, confidence calibration, tool latency

Phase 5: File Findings (2 min)

  1. Write a structured report to docs/dogfood/run-{timestamp}.md
  2. Include: all scores, screenshots paths, specific issues found, recommended fixes
  3. If any score < 3, create a specific fix task description
  4. Compare against previous dogfood run if one exists

Output Format

# Dogfood Run — {date}

## Scores
| Surface | Clarity | Evidence | Actionability | Visual | Speed |
|---------|---------|----------|---------------|--------|-------|
| Decision Workbench | X/5 | X/5 | X/5 | X/5 | X/5 |
| Postmortem | X/5 | X/5 | X/5 | X/5 | X/5 |
| Telemetry | X/5 | X/5 | X/5 | X/5 | X/5 |
| MCP Tools | X/5 | X/5 | X/5 | X/5 | X/5 |

## Issues Found
1. [P0/P1/P2] Description — file:line

## Recommended Fixes
1. Description — expected impact

Read the full file on GitHub · 76 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 · 76 lines · 25 tokens per session scan A 08354f1392d1

Subscribe to this mod's changes

dogfood-loop is an agent published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed 19d ago), licensed MIT. It adds 25 tokens to every session and 778 once invoked, about $0.0001 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.