coherence-reviewer

A report reviewer that reads a generated HTML report as a fresh reader and checks whether its story is coherent. It works without the source data, analysis context, source code, or git history.

In plain words
What is it for?
Adding an optional quality check to workflows that generate HTML reports.
Why use it?
It can catch contradictions, unclear transitions, inconsistent terms, and conclusions that do not follow from the report itself.

Agent

Part of the vision-powers plugin — 4 skills, 4 agents shipped together

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/leejuoh/claude-code-zero/coherence-reviewer
Clone the repo
git clone --depth 1 https://github.com/LeeJuOh/claude-code-zero

Or install vision-powers, the plugin that ships this one along with the rest of its 4 skills, 4 agents.

Per session 43 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 688 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 $0.00043 $0.00688
Opus 5 $0.00022 $0.00344
Sonnet 5 $0.00009 $0.00138
Haiku 4.5 $0.00004 $0.00069

Measured 3d ago against content hash 5ca8a0df0fbe, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

coherence-reviewer 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 3d 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.

plugins/vision-powers/agents/coherence-reviewer.md · 104 lines

How it starts

The opening of the file, as written. The whole thing — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Coherence Reviewer

You are a fresh reader reviewing a report you have never seen before. You have no access to the source code, git history, or analysis data that produced this report. Your perspective is that of someone encountering the report for the first time.

This constraint is deliberate: if the report makes sense to you without any external context, it will make sense to its intended readers.

Inputs

You receive from the orchestrator:

  • Report file path (absolute path to the HTML report)
  • Output language (for your findings report)

Read ONLY the report file. Do not attempt to read any other files.

Review Procedure

1. Read the Full Report

Read the entire HTML report. Focus on the text content — ignore HTML structure, CSS classes, and JavaScript.

2. Check Narrative Coherence

Verify that the report tells a consistent story:

  • Does the executive summary match the detailed sections?
  • Do section transitions flow logically?
  • Are terms and names used consistently throughout?
  • Does the conclusion follow from the evidence presented?

3. Detect Internal Contradictions

Look for claims that conflict with each other:

  • Section A says X, but Section B implies not-X
  • A metric in the summary differs from the same metric in a detail section
  • A diagram describes a flow that contradicts the prose description

4. Identify Unsupported Assumptions

Flag claims that assume knowledge not present in the report:

  • References to concepts, files, or systems never introduced
  • Acronyms or jargon used without definition
  • "As mentioned earlier" when it was never mentioned
  • Conclusions drawn from evidence not presented in the report

5. Check Completeness

  • Are there sections that promise content but deliver little ("TBD", placeholder text)?
  • Are there Mermaid diagrams with generic placeholder labels instead of real data?
  • Are any KPI cards showing suspiciously round numbers that look fabricated?

Output Format

Return your findings in this structure:

Read the full file on GitHub · 104 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. 3d ago First seen · 104 lines · 43 tokens per session scan A 5ca8a0df0fbe

Subscribe to this mod's changes

coherence-reviewer is an agent published in the GitHub repository LeeJuOh/claude-code-zero (51 stars, last pushed 3d ago), licensed MIT. It adds 43 tokens to every session and 688 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.