39 listings reviewed. 11 approvals. A perfect score.
The biggest round yet — 23 sellers in the queue, 13 of them new since round 3. Every seller with a pending listing received a per-listing review email with their score, exact fixes, and realistic ceiling. Re-reviews run within 24 hours of a push, through contest close on 25 August, 23:59 UTC.
| # | Seller | Score | Notes |
|---|---|---|---|
| 1 | dave | 100 | 2 approved listings · 2 confirmed bug bounties (spend-cap, sandbox) |
| 2 | shweta | 93 | best-documented handler in the contest · credential-collision bounty |
| 3 | jayy | 80 | zendesk-integration · biggest three-round climb (26 → 80) |
| 4 | mufeedcreative | 77 | pipedrive-revenueops (active since round 2) |
| 5 | nexicturbo | 76 | cleanest first submission of round 4 |
| 6 | muhammad-akif-janjua | 72 | +30 in one round — biggest single-round climb |
| 7 | imanchuk | 72 | 2 approved · first module_dependency workflow |
| 8 | imtoocompedidiv | 72 | asana-project-ops |
| 9 | tinyops-studio-llc | 70 | github-delivery-operations (active since round 2) |
| 10 | leonard-river | 68 | 25-command QuickBooks module |
Scores are the reviewed listing score plus confirmed bug bounties (+5 per platform bug that ships as a fix, credited in the station release notes). Dave holds two; Shweta holds one.
The first runnable workflow. dave/retainer-billing-run ships an engine_spec — it executes through the governed DAG engine with a declared max_spend_centscap enforced at runtime, saga rollback, and per-node signed receipts. This is the composition pattern the platform is built around, and it's now the reference shape in the docs.
The first declared dependency. imanchuk/incident-airlock declares module_dependency on its companion Cloudflare module — buyers get an install-time "install X first" instead of a mid-run failure.
- · Vault-only auth — env-var credential reads fail review. Every approved module resolves credentials through the vault.
- · mode + risk on every command — now the norm: 14 of 20 reviewed modules ship it complete (round 1: almost none).
- · Storefront depth — the leaders run 1.5–3k-char descriptions with a worked example per command, plus homepage, tests, and video links.
- · Real bug reports — three platform bugs found by contestants have shipped as fixes, each credited in the release notes and worth +5.
- · Max 2 pending submissions per seller — duplicates are soft-removed, latest kept.
- · Slugs must be
seller/name— malformed slugs are rejected at lint. - · Needs-work is a punch list, not a rejection — push fixes and ping in Discord; re-review lands within 24h.
railcall update: the sandbox network fix, the credential_spec collision fix, runnable workflows, and the Runs governance tab all landed since v0.42. Full history at railcall.ai/changelog.