Run the canonical Campaign Receipts production pipeline end-to-end for BOTH programs on the @CampaignReceiptsYoutube channel — SEALED 2016 (145-promise audit) and CR new-news (single-race finance receipts). Spawns one specialist subagent per stage; each specialist's persona IS the gate. No council REVISE loops, no story-score-100 binding gates. Use when the founder says "produce SEALED promise N" / "make a CR new-news piece on RACE" / "make me a short on TOPIC" / "run the pipeline" or after a topic brief lands. Reads eng/PIPELINE-STEPS-AND-OWNERS.md as the canonical step→owner table and .claude/skills/cr-production-pipeline/stages.yaml as the machine-readable stage list.
Use this skill whenever NTO/New Testament Only uses Remotion for episode graphics, quote cards, canon comparisons, timelines, source cards, subscribe cards, or explanatory animations. This skill prevents cheap 1995-game / PowerPoint / moving-board visuals by enforcing fixed documentary boards, internal-only animation, typography hierarchy, visual QC, and an explicit tool-choice gate that rejects Remotion when AI video, licensed stock, or cinematic footage is the better tool.
Design system, methodology, and component vocabulary for the AgentCompanies portfolio (Campaign Receipts, SEALED 2016, EstimateProof, and sibling agent-company sites). Invoke when the user asks to build, redesign, restyle, or extend any marketing page, product surface, share graphic, or component for any of these sites — or to launch a new sibling agent-company site that should match the family aesthetic. Provides the "audit document / paper-receipt" benchmark, drop-in design tokens, type pairing, component anatomies (Receipt, Stamp, StatTile, Leaderboard, MethodCard, ShareTile, etc.), screenshot-share artifact patterns, and the design-exploration workflow (questions → declared system → build → handoff).
Use whenever the agent is about to decide, pick, choose, select, design, write, name, title, thumbnail, hook, copy, pricing, strategy, prompt, photo, music cue, audit, or review anything in a domain where a specialist persona exists. Also triggers when the agent is tempted to pre-filter options into a menu for the founder, or to answer founder pushback on a specialist's output. Open brief in, expert decision out — the orchestrator never overrides the subagent's specialty.
Bot-first support widget for EstimateProof. FAQ auto-deflection + email fallback. Maintains zero-human support rule. Deploy at 10 paying shops. Status: READY (milestone-gated).
Tracked short links via Dub.co for SEALED Press and CampaignReceipts. Know which channels drive conversions. SaaS free tier (1K links/mo). Status: READY — deploy when launching marketing campaigns.
Micro-surveys and structured dispute forms. SEALED /sample intent capture, CampaignReceipts dispute submission with fact-check workflow. Self-hosted on Render. Status: READY — deploy when sites have traffic.
LLM observability for EstimateProof's Claude API calls. Traces every call with cost, latency, tokens. Protects the $100/mo company cap. Self-hosted on Render. Use when checking LLM spend, debugging prompt issues, or setting cost alerts.