The problem is not “Medium with extra steps.” The problem is GHT (and STT skills) piling
up without a written take. RRV fetches GitHub REST data and asks MiniMax for a first-person
developer review. Output is a draft you edit, publish, dismiss, or delete — or
a drip cron run_rrv_drip.py auto-publishes up to auto_publish_max_per_day.
Manual queue: paste URLs / owner/repo, Review Now or Later. Skill-shaped rows (matching
stt_skills.name from run_stt_to_rrv.py) must hit skill generation, not
the repo path — that bug was fixed 2026-08-17. Login for the desk; landing public; cron process
queue is localhost. NLG and ShipLog consume published work. Not GitHub’s blog.
Publish is a human or a capped drip. There is no silent firehose to the public blog.
Not DEV.to. Not a stolen GitHub README as the article.
Skill rows must not use the repo generator. That was a real bug.
auto_publish_max_per_day. If MiniMax is dry, drafts stop.
A published review is factory content, not a vanity URL on this landing.
Will not treat STT skills as GitHub repos. Landing is not the public blog index.
If you wanted raw GHT stars, stay on GHT.
Cards match RRV README.
Login for /rrv/app. This page is not the public blog.
Open RRVRequires a MelTuc account. Create one free.
AI-powered auto-blogging engine that turns GHT trending repos into developer posts. Health stays cheap. If this page disagrees with the signed-in app, trust the app — the product contract and tests enforce that rule.