Socialseal Blueprint Builder

SocialSeal Blueprint Builder

Overview

A SocialSeal blueprint is the compiled, evidence-grounded answer to "what makes content win for this opportunity." It is generated from real exemplar videos, not authored from intuition. It carries best_practices[], evidence[], the selected exemplars, and a set of shot panels that downstream briefs and editor handoffs follow.

This skill drives vnext-blueprints-*. See references/production-pipeline.md for the end-to-end flow and references/mcp-and-cli-usage.md for call patterns. Use socialseal-reference-video-analysis first to select and analyze exemplars.

Generate only after the benchmark direction and analysis depth are approved. A blueprint compiles evidence; it does not replace brand utility facts, asset choices, or visual-direction approval.

When to Use

• Generating a blueprint version for an opportunity scope.
• Reading blueprint history/versions and the latest best_practices/evidence.
• Reading shot-lift rows and pinned shot assets (the panels referenced in the editor handoff coverage table).
• Diagnosing and fixing a missing_data blueprint.

Inputs

Required:

• workspace id and the opportunity opportunityKey
• a scope: topic + pillarId, competitor + competitorBrandIds, tracking_group + trackingGroupId, list + listId + readinessRunId, or manual + videoUids

Good to have:

• refinement: pinnedVideoUids, excludedVideoUids, promotedCandidateTarget, retrievalPrompt (not for manual/list)
• platformIds, region, language, timePeriod
• a title for the blueprint

Workflow

1. Confirm grounded exemplars. Run the preview path in socialseal-reference-video-analysis until the promoted exemplar set is right.
2. Generate the blueprint. Call vnext-blueprints-generate (without previewOnly). The engine queues analysis for any promoted exemplar missing it and writes a version.
3. Interpret status.
   • draft: analysis pending. Poll until ready.
   • generated: ready to use.
   • missing_data: no qualifying evidence. The engine writes an explicit version with a reason (e.g. "no tracking group keywords found for this scope"). Do not invent practices. Fix the scope/keywords/time window and regenerate.
4. Read the blueprint. Use vnext-blueprints-read for the version's best_practices, evidence, and selected exemplars.
5. Read shot-lift. Use vnext-blueprints-shots-read to get shot panels and pinned shot assets (signed URLs). Each panel has a panelId used in the editor handoff coverage table.
6. Refresh shots if stale. Use vnext-blueprints-shots-refresh to requeue shot assets.
7. Hand off. Pass blueprintId (+ version) to socialseal-creator-briefing and the panels to socialseal-asset-planning for a source-clip/editor handoff.
8. Record state. Mark the blueprint generated, draft, or missing_data in the workflow manifest. Never describe a generated blueprint as finished content.

Tool Calls (MCP-first)

Generate (tracking-group scope):

socialseal_call_tool {
  "function": "vnext-blueprints-generate",
  "workspaceId": "<workspace-id>",
  "body": {
    "workspaceId": "<workspace-id>",
    "opportunityKey": "<opportunity-key>",
    "scopeType": "tracking_group",
    "trackingGroupId": <group-id>,
    "timePeriod": "30d",
    "promotedCandidateTarget": 12,
    "title": "<blueprint-title>"
  }
}

Read the latest version and shot-lift:

socialseal_call_tool { "function": "vnext-blueprints-read", "workspaceId": "<workspace-id>", "body": { "opportunityKey": "<opportunity-key>" } }
socialseal_call_tool { "function": "vnext-blueprints-shots-read", "workspaceId": "<workspace-id>", "body": { "blueprintId": "<blueprint-id>", "signedUrlSeconds": 3600 } }

CLI equivalents:

npx -y @socialseal/cli tools call --function vnext-blueprints-generate --workspace-id <workspace-id> --body @blueprint.json --pretty
npx -y @socialseal/cli tools call --function vnext-blueprints-read --workspace-id <workspace-id> --body '{"opportunityKey":"<opportunity-key>"}' --pretty
npx -y @socialseal/cli tools call --function vnext-blueprints-shots-read --workspace-id <workspace-id> --body '{"blueprintId":"<blueprint-id>"}' --pretty

Output

• blueprintId and version, with status
• best_practices[] each tied to evidence[]; when surfacing a practice to a person, cite the exemplar by video title/URL and @handle and keep the video_uid as a traceability note
• selected exemplars with scores and matched keywords

A compiled blueprint raises confidence because its practices are grounded in multiple surfacing exemplars, but it is still an indicative pattern, not proof a given execution will perform. Frame downstream bets as testable. See references/evidence-and-confidence.md.

• shot panels (panelId, shot label, kind) and pinned shot assets
• a clear missing_data note and remediation when applicable

Do / Don't

Do:

• ground every best practice in cited exemplar evidence
• reuse one opportunityKey across blueprint, brief, and editor handoff
• use panelIds to identify shots in the handoff coverage table; this does not create persisted SocialSeal mappings
• record the blueprintId and version for downstream skills

Don't:

• author best practices when the engine returns missing_data
• mix scopes in one blueprint to force a result
• edit shot panels by hand outside the engine
• expose literal workspace/blueprint IDs in shared artifacts

Troubleshooting

• missing_data: widen keywords/scope, adjust timePeriod, or switch scope type; for competitor scope ensure competitorBrandIds resolve to active aliases.
• Stuck in draft: promoted exemplars are still analyzing; poll vnext-blueprints-read and video_analysis status.
• List scope errors (READINESS_STALE/READINESS_BLOCKED/READINESS_WARNING_REQUIRES_OVERRIDE): re-run list readiness; pass allowWarningOverride only when intentional.
• Empty shot-lift: run vnext-blueprints-shots-refresh, then re-read.

Verification Checklist

• Blueprint reached generated (or missing_data was surfaced and addressed).
• Best practices are grounded in cited exemplar evidence.
• Shot panels and panelIds are recorded.
• blueprintId, version, and opportunityKey are handed to brief/asset skills.