DiffSentry — Risk summaries for agent-written PRs

Flag the riskiest logic in AI-generated changes before the review queue explodes

Problem

Developers report review queues backed up and diffs growing, with nobody confident the code is safe to ship; agent-generated code moves fast but the review process didn't evolve, so reviewers still mentally simulate what the agent did and hope the happy path is covered.

Solution

A PR bot that connects to GitHub/GitLab, identifies which changes came from coding agents, and highlights high-risk patterns such as unhandled exceptions, missing edge cases, permission bypasses and concurrency issues; it also marks code blocks with no corresponding test.

Users

Engineering teams using AI coding, tech leads who need fast sign-off, indie developers.

MVP

GitHub App integration; detect metadata of agent-generated commits; produce a risk checklist and missing-test hints per new/changed function; post as a PR comment; free up to 5 PRs/month, then subscription.

Revenue

Monthly subscription by repo or seat; free for open-source repositories.

Why now

Agent-generated code volume keeps growing while review automation lags badly; an indie developer can enter cheaply through the GitHub App ecosystem without building a code host.