Decision
Include the Eddie editor agent in the JOLT V1 build from the start rather than retrofitting it later. Build lightweight versions of all agents first, then add Eddie before or shortly after launch. Content DNA agent should auto-generate a branded strategy PDF from kickoff call transcript plus onboarding form.
Context
JOLT agent architecture review revealed a key gap: no Eddie editor agent included. Without Eddie, agents will hallucinate or forget client-specific details. Megaphone precedent: Casper improved significantly once Eddie was added. Eddie is low cost relative to Trent and Connie agents.
Options considered
- Retrofit Eddie after initial build: rejected because fixing architectural gaps post-launch is harder and risks client-specific hallucinations from day one
- Skip Eddie entirely: rejected based on Megaphone/Casper precedent
Consequences
- Tatem’s input required to train Eddie on best scripting practices
- Frame Eddie work to Tatem as making her life easier and giving her ownership of the agent
- If Tatem is hesitant about proprietary info, connect her with Blake (who can speak to how Eddie improved Casper)
- TikTok trend scraper: build out, test varying post counts (25 may not be enough), find point of diminishing returns to determine cost-per-request at scale
- JOLT accounts (Claude platform + Apify) created separately from Megaphone; billed to Rye Valley Amex