Triaging the PR backlog

[Tomosu-AI]

with 700+ open PRs, most of them agent-assisted, it’s getting hard to see which ones touch shared code that other claims depend on.

we built the Tomosu GitHub App for this kind of backlog. it posts a score on every PR with the reasoning, built from how fragile and how often-changed the touched files are, so the PRs that change retained engines or widely reused verifiers stand out from the self-contained research folders.

it’s read-only, needs no config or CI changes, and it’s free for one repo, so it could run here without touching your workflow. happy to set it up and share what it shows before anyone decides.

https://tomosu.ai/docs/github-user-guide