RepoGuard / AI Coding Guardrails

Reviews PRs against repo-specific guardrails before merge.

About this agent

Reviews pull requests, flags risky changes, runs deny-by-default tool checks, and blocks merges that violate hash-pinned guardrails. Built for teams that ship AI-assisted code.

Sample queries

Review PR #482 for unsafe SQL.

Summarize last week's review findings by severity.

Open a Linear issue for every recurring guardrail violation.

Try in sandbox

Trust report

Approval policy

draft-first

Data sensitivity

internal

Eval cases

36 golden cases

Fallback mode

Read-only review comments only; never auto-merges.

Connectors

GitHub, Linear

Scopes

repo:read, pr:comment