Communitygithub.com

simpletibr/simplicio-loop-marketing

Assembles a video generation prompt by picking the provider via PROVIDERS.md and delegating to the matching provider-specific builder

Was ist simplicio-loop-marketing?

simplicio-loop-marketing is a Claude Code agent skill that assembles a video generation prompt by picking the provider via PROVIDERS.md and delegating to the matching provider-specific builder.

Funktioniert mit~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/simpletibr/simplicio-loop-marketing/tree/HEAD/.skills/video-prompt-builder

In Ihrer bevorzugten KI fragen

Öffnet einen neuen Chat, in dem dieser Agent-Skill bereits geladen ist.

Dokumentation

Video Prompt Builder

Generic, provider-agnostic dispatcher for video generation prompts. Reads the routing matrix in .specs/architecture/PROVIDERS.md, decides which provider should run the piece, and delegates to the right specialist (higgsfield-prompt-builder, topview-prompt-builder, wavespeed-batch, hyperframes-prompt-builder).

When to invoke

  • Any video generation step in the pipeline (brief -> script -> creative when creative type is video).
  • When a piece declares type: video in .specs/pieces/*.md.
  • A/B testing video hooks across providers and the orchestrator needs a single entry point.
  • Re-running the same brief on a different provider after an A/B winner is picked.
  • When an upstream agent has a brief but does not know which provider to choose.

Inputs

  • brief: object. Visual brief with at least subject, motion, mood, aspect, duration_seconds.
  • task_kind: string. One of cinematic_reel, motion_control, ugc_product_holder, product_demo, talking_head, batch_hook_test, motion_typography, data_viz_reel, programmatic_short.
  • piece_path: string, optional. Path to the piece file. May contain provider_override.video.
  • dry_run: boolean, optional. When true, returns the resolved provider and parameters without calling the MCP.

Process

  1. Read .specs/architecture/PROVIDERS.md and parse the Video matrix into a task_kind -> provider map.
  2. If piece_path is set and contains provider_override.video, use the override.
  3. Else look up task_kind in the matrix. If unknown, fall back to the matrix default for cinematic_reel.
  4. Verify the chosen provider has the env vars set (HIGGSFIELD_MCP_ACTIVE, TOPVIEW_API_KEY, WAVESPEED_API_KEY, HYPERFRAMES_ACTIVE).
  5. Map provider to specialist skill: Higgsfield -> higgsfield-prompt-builder, Topview -> topview-prompt-builder, Wavespeed -> wavespeed-batch, Hyperframes -> hyperframes-prompt-builder.
  6. Translate brief fields into the input shape the specialist expects.
  7. Invoke the specialist and capture its params and mcp_tool output.
  8. If dry_run, return the resolved provider, the params, and the MCP tool. Otherwise call the MCP and return the artifact reference.
  9. Log the run to data/video-usage.jsonl with timestamp, provider, task_kind, and outcome.

Outputs

  • provider_used: string.
  • params: object. Final parameter object ready for the MCP call.
  • mcp_tool: string. The MCP tool name.
  • artifact_ref: string, when not in dry_run. Path or URL to the generated video.
  • cost_estimate_usd: number.

Examples

Example 1: cinematic reel routed to Higgsfield

Input: { brief: { subject: "model in linen suit", motion: "tracking shot", mood: "editorial", aspect: "9:16", duration_seconds: 6 }, task_kind: "cinematic_reel" } Output: { provider_used: "higgsfield", mcp_tool: "higgsfield_seedance_generate", params: {...} }

Example 2: UGC ad routed to Topview

Input: { brief: { subject: "skincare bottle held by avatar", motion: "static", mood: "friendly", aspect: "9:16", duration_seconds: 15 }, task_kind: "ugc_product_holder" } Output: { provider_used: "topview", mcp_tool: "topview_avatar_generate", params: {...} }

Example 3: weekly KPI reel routed to Hyperframes

Input: { brief: { headline: "Semana 21 em números", aspect: "9:16", duration_seconds: 12, variables: { leads: 42, delta_pct: 12 } }, task_kind: "programmatic_short" } Output: { provider_used: "hyperframes", mcp_tool: "hyperframes_render", params: { composition_spec: {...}, render_args: {...} } } — local render via the hyperframes + hyperframes-cli skills.

Failure modes

  • task_kind not in matrix: log a warning and use the cinematic_reel default.
  • Required env var missing for the chosen provider: drop to the next provider in the matrix and surface a warning.
  • Specialist skill returns an error: surface it and do not auto-retry on a different provider unless the caller opts in.
  • Duration exceeds provider limit: trim to the provider max and surface a notice.

Related skills

  • higgsfield-prompt-builder: specialist for Soul, DoP, Seedance.
  • topview-prompt-builder: specialist for UGC and avatar-driven ads.
  • wavespeed-batch: specialist for cheap batch hook tests.
  • hyperframes-prompt-builder: specialist for HTML-rendered motion typography, data-viz reels, and parametrized programmatic shorts.
  • llm-router: separate dispatcher for text generation; same pattern, different domain.
  • qa-tech-specs: validates the final video against the platform spec.

Individual skills in this repo

This repo contains 5 individual skills — each has its own dedicated page.

Verwandte Skills