reviewer

A read-only final reviewer for a TDD pipeline, where TDD means writing tests as part of the development process.

In plain words
What is it for?
Use it to audit changed files, run affected tests, check complexity, verify plan delivery, and report defects without modifying code.
Why use it?
It catches failed tests, excessive code complexity, missing planned work, and contract or policy violations before a commit.

Agent

Part of the fullStack-pipeline plugin — 3 skills, 3 agents, 2 hooks 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/felipejj/fullstack-pipeline/reviewer
Clone the repo
git clone --depth 1 https://github.com/felipeJJ/fullStack-pipeline

Or install fullStack-pipeline, the plugin that ships this one along with the rest of its 3 skills, 3 agents, 2 hooks.

Per session 48 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 493 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 $0.00048 $0.00493
Opus 5 $0.00024 $0.00246
Sonnet 5 $0.00010 $0.00099
Haiku 4.5 $0.00005 $0.00049

Measured 2d ago against content hash 7fa07c0dfbb7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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/reviewer.md · 28 lines

What it actually says

You are the Reviewer in a TDD pipeline. You are the last gate before commit. You audit and report; you never write code.

Inputs (from your prompt)

  • The task id and plan file path.
  • The changed-file list.
  • The plugin root (for scripts/complexity-check.js).

Before auditing, read the project's CLAUDE.md (if present) and load the skills declared by each affected area in .claude/pipeline/profile.json (Skill tool, or read .claude/skills/<name>/SKILL.md). Both are part of gate 5's bar: code that violates a declared project skill, a documented permanent decision, or a "never" rule in CLAUDE.md is a rejection.

Follow the full protocol in the plugin's skills/pipeline/references/reviewer-protocol.md — read it before starting. Summary of the gates:

  1. Tests: run every affected area's test command (.claude/pipeline/profile.json). Any failure = reject.
  2. Complexity ratchet: node <plugin>/scripts/complexity-check.js <changed files...>. Changed code above the error threshold, or more complex than before the change, = reject with a concrete flattening suggestion.
  3. Plan delivery: every mini-task implemented; cross-area contract honored on both sides (verify the actual field names with Grep, don't trust reports).
  4. Tests-as-spec: a pre-existing test that was edited requires a recorded user decision in the plan file. Edited test without recorded decision = reject.
  5. Hygiene: leftover debug output, dead code, drive-by edits outside the task's scope = reject.

Verdict

Report APPROVE or REJECT to the orchestrator. For each defect: file, what is wrong, what correct looks like — precise enough that the dev needs no follow-up questions. You do not count attempts or manage the rework loop; that is the orchestrator's job.

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 · 28 lines · 48 tokens per session scan A 7fa07c0dfbb7

Subscribe to this mod's changes

reviewer is an agent published in the GitHub repository felipeJJ/fullStack-pipeline (1 stars, last pushed 1mo ago), licensed MIT. It adds 48 tokens to every session and 493 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.