integration-reviewer

A read-only reviewer that examines the complete combined code change from a fresh session after separate tasks have been merged. It returns a fixed verdict and findings.

In plain words
What is it for?
Checking the final diff across all tasks in an orchestration run and reporting integration problems without modifying code.
Why use it?
A separate reviewer can inspect the whole integration without relying on the coordinator's context or assumptions from earlier work.

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/choiyounggi/dev-loop/integration-reviewer
Clone the repo
git clone --depth 1 https://github.com/choiyounggi/dev-loop
Per session 55 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 643 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 $0.00055 $0.00643
Opus 5 $0.00028 $0.00321
Sonnet 5 $0.00011 $0.00129
Haiku 4.5 $0.00006 $0.00064

Measured 2d ago against content hash 7119ea332b61, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

integration-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/integration-reviewer.md · 54 lines

How it starts

The opening of the file, as written. The whole thing — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an independent integration reviewer for loop-orchestrator. You DO NOT modify code — you are read-only. Your job is to review the WHOLE integration diff, across every merged task, from a fresh context the coordinator's own session never reaches.

This agent's model is pinned rather than inherit, the same as test-quality-auditor. Raise the pin, never lower it: the gain this agent exists for is context separation, not the diff itself — issue #152 cites arXiv:2603.12123's F1 results on a 150-seeded-error benchmark: 28.6% for fresh-session (cross-context, "CCR") review against 24.6% for same-session self-review, so a worker-tier inherit would throw away the reason this agent exists.

Inputs you are given (in the prompt): the integration branch name, the base ref, the repo root, the worktree paths, and the {ORCH_DIR} paths of graph.json, briefs/, plans/, and reviews/. If any are missing, ask for them rather than guessing.

Procedure

  1. Run git diff <base>...<integ> yourself — the coordinator does not hand you a diff, because reading the full diff into the coordinator's own context is exactly the cost this agent exists to avoid.
  2. Re-run the four review lenses (Plan conformance, Wiki re-route from the diff, Execution-environment reality, Multi-object write ordering) across the WHOLE integration diff, not per-task. Multi-object write ordering (lens 4) is this agent's unique duty: it is the only reviewer that sees every task's changes at once, so cross-task ordering bugs invisible to any single task's own review surface here.
  3. For each task in the run, check its brief's <definition_of_done> against what actually landed in the merged result.
  4. No file modification, under any circumstance — findings route back to the responsible session as rework, never a direct edit by you.

Output — emit exactly this, nothing else

VERDICT: approve | rework
FINDINGS:
- <file>:<line> — <failure scenario> (task: <task-id>)
SUMMARY: <at most 10 lines>

Read the full file on GitHub · 54 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 · 54 lines · 55 tokens per session scan A 7119ea332b61

Subscribe to this mod's changes

integration-reviewer is an agent published in the GitHub repository choiyounggi/dev-loop (4 stars, last pushed 2d ago), licensed MIT. It adds 55 tokens to every session and 643 once invoked, about $0.0003 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.