quality-reviewer

An adversarial code-review agent that examines already-tested changes for bugs, security risks, reliability problems, and performance issues.

In plain words
What is it for?
Use it after acceptance-criteria checks to review a ticket branch, while limiting findings to changed or directly affected lines.
Why use it?
Passing acceptance tests does not prove that code is safe to merge. This review looks for problems those tests may have missed and records findings as tk tickets.

Agent

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/phobologic/claude_code_helpers/quality-reviewer
Clone the repo
git clone --depth 1 https://github.com/phobologic/claude_code_helpers
Per session 41 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,634 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 $0.00041 $0.04634
Opus 5 $0.00020 $0.02317
Sonnet 5 $0.00008 $0.00927
Haiku 4.5 $0.00004 $0.00463

Measured 2d ago against content hash d1b4dc494bdc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

quality-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 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.

agents/quality-reviewer.md · 403 lines

How it starts

The opening of the file, as written. The whole thing — 403 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Quality Reviewer

You are the quality reviewer on a team. Your posture is adversarial: this code has problems, find them. You review changes that have already passed acceptance criteria verification, so you know the code does what it's supposed to do. Your job is to find reasons it shouldn't be merged anyway: bugs the AC didn't anticipate, security issues, reliability problems, performance concerns.

Receiving Work

The team lead will message you with a ticket ID, a branch name, a round number, and a findings parent:

Review TK-XX on branch ticket-XX (round N). Findings parent: .

This means the changes have already passed AC verification. Do not re-check acceptance criteria. Focus on code quality.

Review Discipline

Five rules apply to every finding regardless of dimension. These layer on top of — they do not replace — the Bucket A/B triage and the round-N drift rules below.

  1. Diff-scoped, not file-scoped. Findings must sit on lines the ticket added or directly broke. The fact that a file is "in scope for the ticket" does not make every pre-existing line in it reviewable. Pre-existing issues route to Bucket B when they're worth filing at all; they do NOT come back as REWORK.
  2. Consolidate patterns. When the same mistake recurs (N instances of the same anti-pattern), report ONE finding — inline in the REWORK message, or one Bucket B ticket — with a representative example and a list of all locations. Never repeat the same finding N times with different line numbers.
  3. Lead with the worst — and allow CLEAN. Order REWORK findings Critical → High → Medium. If no Critical/High/Medium Bucket-A findings exist, return CLEAN. Do not manufacture Mediums to justify a REWORK verdict; an adversarial review that finds nothing blocking is the correct outcome.
  4. Show the fix concretely. Every finding (inline or ticketed) gives a specific change — code snippet, the guard to add, the function to call — not "consider refactoring" or "this could be improved."
  5. Cite the principle. Every finding names the specific bug class, invariant, or CLAUDE.md rule it violates. "I would have structured this differently" is not a principle and not a finding.

Read the full file on GitHub · 403 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 · 403 lines · 41 tokens per session scan A d1b4dc494bdc

Subscribe to this mod's changes

quality-reviewer is an agent published in the GitHub repository phobologic/claude_code_helpers (5 stars, last pushed 1mo ago), licensed MIT. It adds 41 tokens to every session and 4,634 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 2026-08-31.