Meta Live Campaign Launch Operator

Use this skill when a Meta campaign already exists or has just been published and the user asks to make live changes, such as correcting URLs, changing headlines, adding another video ad set, setting end dates, activating delivery, or confirming that live objects are configured correctly. This skill turns the VAN Studios workflow into a repeatable campaign-operations process.

This skill complements meta-safe-campaign-builder. Use meta-safe-campaign-builder for first-time paused creation. Use this skill when the task involves live updates or approved activation.

Non-Negotiable Safety Rule

Never publish, activate, edit URLs, change budgets, pause delivery, or create new active delivery objects unless the user has explicitly approved that live action in the current task. If the user says they are happy for you to publish or activate, treat that as approval for the specified objects only. Do not infer approval for additional campaigns, budgets, or unrelated assets.

Keep access tokens, ad account IDs, and raw API credentials out of user-facing reports unless the user already provided those IDs and they are necessary operational IDs. Never include tokens in Markdown handovers.

Required Inputs

Confirm these before write calls. If one is missing and cannot be recovered from existing campaign objects, ask the user or perform read-only inspection first.

Standard Workflow

1. Load Dependencies and Local Instructions

Read this skill, then read meta-safe-campaign-builder and meta-paid-ads-challenge-operator when the campaign is a challenge, webinar, paid-ticket event, Funnel Junkies campaign, or Van Studios campaign. Check for local AGENTS.md instructions before editing files or scripts.

2. Run a Read-Only Preflight

Before making changes, inspect the target campaign, ad set, ad, and creative. Save a sanitized JSON preflight file. Confirm the campaign objective, budget level, current statuses, promoted object, targeting, attribution, existing creative URL, existing headline, and whether the campaign is already live or in review.

Use read-only calls first. If the preflight shows the wrong campaign, wrong account, wrong pixel, or unexpected budget, stop and ask for confirmation.

3. Prepare a Change Plan

Create a short implementation plan that separates confirmed facts from intended changes. The plan should state which existing ad will be updated, what new ad set/ad will be created, the URL, headline, budget, event end date, and whether objects will be created paused or active.

For live updates, explicitly record the user’s approval phrase or request in the script output as intent: LIVE_UPDATE_APPROVED_BY_USER or equivalent.

4. Upload New Video and Thumbnail

For MOV/MP4 files, use chunked video upload through graph-video.facebook.com. Extract a thumbnail frame with ffmpeg, upload it to /act_{ad_account_id}/adimages, and use the returned image_hash inside video_data. Poll video status using safe fields only: id, title, status, and created_time.

If the video is still processing, wait and recheck. Do not create duplicate videos unless the upload failed before returning a video ID.

5. Create or Update Creatives

Meta ad creative content is not reliably editable in place. Prefer creating a new creative with the corrected URL, headline, copy, CTA, video ID, and thumbnail hash, then attach that creative to the existing ad. This preserves traceability and avoids partial creative edits.

When adding a second video test, create a separate creative and ad, even if the copy and URL match the first creative. This makes ad-level reporting clean.

6. Create Additional Ad Set and Ad

When adding a new ad set inside the same campaign, mirror the existing ad set’s proven structure unless the user requested a different audience. Reuse targeting, attribution, optimization event, bid strategy, and promoted object. Apply the requested daily budget and end date.

For event campaigns, if the user says the campaign should finish on the event start day, set the end time to the end of that day in the account’s relevant timezone unless they explicitly request a morning cutoff. This avoids cutting delivery off at the beginning of the event day.

7. Activate Only Approved Objects

If approval is active, set the campaign, relevant ad sets, and relevant ads to ACTIVE. If approval is not active, create all new delivery objects as PAUSED and hand over launch instructions.

Do not activate unrelated ad sets or legacy ads inside the same campaign. Activate only the objects scoped in the user’s request.

8. Run Final QA

Run individual or batch read calls after the write calls. Confirm:

If Meta returns a rate-limit error after write calls, stop non-essential API calls, save the partial result, and retry later with a lightweight batch QA. Do not keep hammering the API.

9. Deliver a Handover

Write a Markdown handover with the final object map, statuses, URLs, budgets, end dates, QA results, any rate-limit or review notes, where to find the ads in Ads Manager, and the next monitoring checklist. Attach the handover and sanitized JSON files.

Rate Limit and Partial Success Rules

Meta can accept write calls and then rate-limit broad list or QA calls. If this happens, read the saved result JSON before retrying. Determine which objects were already created, then run the smallest possible QA query. Do not rerun the full creation script unless it is idempotent and checks for existing object IDs.

Performance-Marketing Guardrails

Optimize for qualified purchases and downstream business outcomes, not cheap clicks. After launch, monitor cost per purchase, landing-page-view rate, checkout completion, buyer quality, attendance, booking rate, and ascension quality. For challenge campaigns, the first 24 hours are a controlled learning window, not a scale window.

Bundled Resources

Use scripts/meta_live_update_from_config.py for a reusable config-driven implementation pattern. Use templates/live_update_config.example.json to create a campaign-specific config file. Use templates/live_update_handover_template.md when writing the final handover.

Completion Criteria

A task using this skill is complete when the requested live changes are either fully applied and QA-confirmed, or a specific blocker is documented with object IDs, partial-success status, and the next safe recovery step. The final response must include the key object IDs and attach the handover plus sanitized QA files.