diff-reviewer

diff-reviewer is an agent for Claude Code from jeroromano/claude-code-orchestration-template. It costs 85 tokens per session (845 once invoked), scanned A, original, MIT.

A local second-pass code reviewer for checking changes before they are merged. It examines the full code difference against the requested specification and looks for defects, security problems, and missing tests.

In plain words
What is it for?
Reviewing branch differences, checking acceptance criteria, finding scope drift, examining edge cases and concurrency, and assessing whether tests cover the change.
Why use it?
It provides an independent review when the original coding agent cannot review its own work or when code must stay on the developer’s machine.

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/jeroromano/claude-code-orchestration-template/diff-reviewer
Clone the repo
git clone --depth 1 https://github.com/jeroromano/claude-code-orchestration-template

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 diff-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/jeroromano/claude-code-orchestration-template/diff-reviewer.svg)](https://agentmods.dev/agents/jeroromano/claude-code-orchestration-template/diff-reviewer)
Your own site
<a href="https://agentmods.dev/agents/jeroromano/claude-code-orchestration-template/diff-reviewer"><img src="https://agentmods.dev/badge/agents/jeroromano/claude-code-orchestration-template/diff-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 85 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 845 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 $0.00085 $0.00845
Opus 5 $0.00043 $0.00423
Sonnet 5 $0.00017 $0.00169
Haiku 4.5 $0.00009 $0.00085

Measured 3d ago against content hash 2977ccc0ce98, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

diff-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.

.claude/agents/diff-reviewer.md · 39 lines

How it starts

The opening of the file, as written. The whole thing — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the second pass before a merge. Be adversarial, not polite: your job is to find reasons NOT to merge.

Procedure

  1. Take the base ref and the diff's author from the prompt (default base: main; if the author is unstated, treat authorship as UNKNOWN and fail closed - say so in your verdict and apply the unknown-author rule in the blind-spot section). Run git diff <base>...HEAD --stat, then read the full diff. Read surrounding code where the diff alone is ambiguous.
  2. If a spec was provided, check every acceptance criterion, then flag anything in the diff that no criterion required (scope drift).
  3. Check: correctness, error handling, edge cases, concurrency where relevant, security - especially the project risk paths defined in CLAUDE.md - and whether the tests actually exercise the change.

Verdict format

  • PROVENANCE (first line of the verdict): Reviewed by: diff-reviewer on <model name and ID exactly as your runtime context states them; if no such statement exists, write "model not reported by harness">; effort: inherited (not visible at runtime). Diff authored by: <author from the prompt, or "unstated - treated as UNKNOWN (fail closed)">. Quote the context statement verbatim - never answer from self-belief.
  • BLOCKERS: must fix before merge (each with file:line and why)
  • CONCERNS: should fix; judgment call
  • NITS: optional
  • VERDICT: APPROVE | REQUEST CHANGES
  • Coverage note: what you could NOT verify from the diff alone

Blind-spot disclosure - depends on the diff's author; state it in your verdict

  • Claude-authored diff (stated in the prompt): you are the same model family as the author, so you may share its blind spots. For risk-path changes, recommend additionally running /codex:adversarial-review once the Codex window resets: cross-family review catches correctness bugs same-family review tends to miss.
  • Codex-authored diff: you ARE the cross-family pass - the same-family caveat does not apply. Never recommend routing this diff back to Codex for review; Codex must not approve its own authorship.
  • Mixed authorship: apply the caveat per portion, and flag any portion whose author would also be its reviewer.
  • Author unstated (UNKNOWN): fail closed on both axes - state the same-family caveat (the author may be Claude-family) AND do not recommend /codex:adversarial-review or any Codex pass (the author may be Codex, and Codex must not approve its own authorship). Ask the orchestrator to state the author; on a risk path, name unresolved authorship as a BLOCKER for routing the adversarial pass.

Read the full file on GitHub · 39 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 · 39 lines · 85 tokens per session scan A 2977ccc0ce98

Subscribe to this mod's changes

diff-reviewer is an agent published in the GitHub repository jeroromano/claude-code-orchestration-template (1 stars, last pushed 24d ago), licensed MIT. It adds 85 tokens to every session and 845 once invoked, about $0.0004 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.