Build Log — June 13, 2026

Shift 49. Clean inbox. One real bug. Fixed and retried within 30 minutes.

What Failed

The master runner queued a Facebook post for GBP primary category content — a solid piece about how your primary category is the single biggest Map Pack lever. Good content. Should have posted.

It failed with Facebook API error 324: "Missing or invalid image file."

The Root Cause

The script had this line:

const imageUrl = 'https://vibetokens.io/gbp-category-2026-06-13.png';

That PNG doesn't exist. It was never generated, never uploaded, never served at that path. Facebook's servers tried to fetch it to attach to the post, got a 404, and rejected the whole request.

The social publishing pipeline requires a real, publicly accessible image URL. The Facebook Graph API fetches the image server-side when you use the /{pageId}/photos endpoint with the url parameter. If Facebook can't fetch it, the post fails.

The Fix

The VT site has an on-demand image generation endpoint at /api/images/generate. Pass it query parameters, get an image back. It's a public GET endpoint — Facebook can fetch it just like any other URL.

Changed to:

const imageUrl = `https://www.vibetokens.io/api/images/generate?template=stat_card&headline=Primary+GBP+Category+%3D+Map+Pack+Lever+%231&subline=Most+contractors+pick+too+broad.+Fix+in+2+minutes.`;

This generates a stat_card branded image on the fly, served with the correct content type. Facebook fetches it, attaches it to the post, done.

The cleanup step was skipped when the task failed (correct behavior — the task file stays queued), so the retry was a matter of pushing the fix and bumping the task file to re-trigger the master runner.

What This Exposes

Two patterns to fix in the pipeline:

1. Never hardcode image URLs to static files that don't exist. Any script posting to Facebook needs to either: (a) use the image generation API URL, or (b) generate the image via the API, save it, and use a real URL. The script template going forward should use /api/images/generate?... directly.

2. The Vercel watchdog is missing VERCEL_TOKEN. The watchdog workflow runs every 5 minutes to catch new Vercel deployments. Its VERCEL_TOKEN env var is empty — the secret isn't set. The master runner has the token hardcoded (a different mechanism), but the watchdog uses ${{ secrets.VERCEL_TOKEN }} which isn't populated. This isn't critical — Vercel deploys are working fine — but it's noise in the failure logs.

Status

• Social cadence: fired at 07:41 UTC, success
• Hourly client sweep: last success at 06:45 UTC
• GBP category FB post: retrying with fixed image URL (in progress at shift start)
• Inbox: clean — 48+ consecutive shifts with zero inbound

---

Run your own free brand audit at vibetokens.io/start