Forest — Brand Films Agent

Get Forest to functioning.

Forest is a six-agent system that runs BlueForest's cold outreach — from first prospect research to a signed brand film deal. This page is the visual overview: the team, the pipeline they run, and where the build stands today.

The team The pipeline

Phase 1 Milestones

0% 0 of 14 complete
  • Review & tune foundation promptsDone
  • Voice calibration locked (house style + bio)Done
  • Forest's SOUL.md reviewed & lockedDone
  • Tech decisions resolved (SDK, deploy, tools)Done
  • North-star goal & metrics definedDone
  • Director Partner Network + Reference Films seededDone
  • Smartlead → HubSpot native integration liveDone
  • Smartlead.ai live + warming completeDone
  • Runtime decision: Claude Code on dedicated Mac Mini (SDK archived)Done
  • Migrate agent to Claude Code skills (host setup + skill conversion)Us
  • Daily whiteboard sheet builtForest
  • 1–2 backup sending domains warmingYou
  • blueforest.studio lean v1 website liveUs
  • First end-to-end test: 5 prospects processedUs

The team

Forest leads five sub-agents. Each has a specialty. Everyone reports to one person.

6 agents
1 human · 1 orchestrator · 5 specialists
1 gate
Reviewer — last line before send

Forest is a team of six agents, not one. Each sub-agent has a single job and a specific voice. Forest is the only one who orchestrates — sub-agents never call each other directly. Reviewer is the last gate: every external draft passes through Reviewer before it reaches Ammon or the outside world.

Ammon Ehrisman, founder of BlueForest Studios
Ammon
founder · final decision-maker
Forest
orchestrator · synthesizer
Runs the nightly loop. Delegates work to sub-agents and reviews what comes back. Produces the morning briefing. Never does sub-agent work himself.
SOUL.md voice — inherited by every sub-agent
Scout
prospect research
journalist-instincts
Produces prospects/{slug}/brief.md
Curator
reference films + director network
film-school-taste
Maintains Director Partner Network sheet · per-film deep-dives
Scribe
outbound in Ammon's voice
craftsman-writing
Produces team/scribe/drafts/
Analyst
funnel · metrics · experiments
skeptical-analyst
Produces daily / weekly / monthly reports
QA Gate
Reviewer
last line before external send
line-editor-directness
Returns verdict + notes per draft
Shared context — load-bearing documents merged into the relevant agent's system prompt at runtime
forest/prompts/SOUL.md Voice baseline — inherited by every agent
forest/prompts/house-style.md Outbound mechanics — Scribe + Reviewer
assets/bio/ammon-founder-story.md Ammon's story, voice, beliefs, process — Scribe
assets/fit-signals.md ICP rubric (Step 0 on every prospect) — Scout
team/forest/learnings.md Institutional memory — Forest, referenced by all

The pipeline

What Forest actually runs, per prospect. Eight stages, top to bottom — from a name to a signed deal.

8 stages
Name → signed deal
Most exit at 2
Tier 3 / anti-signal stops here — by design

Eight stages, top to bottom. The main path is what a Tier 1 prospect looks like end-to-end. Branches kick out at every decision point — most prospects exit early, which is correct. Quality of contact trumps quantity. Analyst observes the whole pipeline from the right rail without intervening.

1
Surface
Incoming signal
A name arrives — Ammon flags one, Scout's prospecting queue produces one, or a prospect replies to a post.
Trigger: name + any seed context
2
Research
Scout
Step 0 fit-signals pass first, then a full brief if the prospect is Tier 1 or Tier 2.
Produces prospects/{slug}/brief.md
Tier 3 / anti-signal
Filed with a short note. Pipeline stops here.
3
Match
Curator
Picks 1–2 reference films from the verified library and a candidate director from the Partner Network.
Appended to the prospect brief
4
Draft
Scribe
Writes the cold email touch 1 in Ammon's voice — FirstName, opener, ≥2 of 3 tonal signatures, real numbers no hedging.
Produces team/scribe/drafts/...md
5
Review
Reviewer · QA gate
Applies the §3.1 checklist — structural, tonal, banned phrases, fact-check — and returns a verdict.
Logs to team/reviewer/logs/{date}.md
Send back
Returns to Scribe with the named gap.
Block
Surfaces to Ammon immediately — fabricated fact or hard rule broken.
6
Approve & send
Forest + Smartlead
Forest queues the approved draft. At Tier 1, Ammon clicks approve. Smartlead sends from the warmed domain. The send is logged to HubSpot via the native integration.
HubSpot contact timeline + Smartlead campaign log
7
Triage reply
Forest
Forest reads inbound via Gmail and classifies each reply: warm, objection, cold, auto-reply, or bounce.
Warm replies → top of tomorrow's morning briefing
No reply
Touch 2 / 3 / 4 scheduled. Each adds something new — never "just bumping this."
8
Conversation → call → proposal → signed
Ammon + Scribe (Reviewer-gated)
A warm reply becomes a discovery call. The call becomes a creative brief + proposal (Scribe drafts, Reviewer gates). The proposal becomes a signed brand film deal.
Goal state · north star: 1 signed deal by July 30, 2026

Decisions log

What's been settled. The full detail behind each lives in the prompt files, the learnings log, and project memory — this is the at-a-glance record.

Phase 1
Foundation review — complete

Runtime & technical

Runtime: Claude Code skills on a dedicated Mac Mini, under Ammon's Max subscription. The earlier standalone-SDK plan (TypeScript, Cloudflare) was built, smoke-tested, then archived.
2026-05-08
Cold email volume: low — 25–50 per week. Quality of contact over quantity.
2026-05-02

North star & offer

North star: 1 signed brand film deal by July 30, 2026. Supporting milestones: 5 cold sends/day, 1 discovery call/week by July 5.
2026-04-24
Offer: a flat $30K per film (travel billed separately). Deliberately a single price — no tiers.
2026-05-16

Voice

Bio pack populated — §2 origin, §3 beliefs, §4 voice, §7 process — all from Ammon's own input.
2026-05-04
Voice locked: the FirstName, opener, "let me know," "by the way," and the disarming-transparency + generous-offer + real-numbers triad. Banned phrases set (synergy, leverage, value-add, game-changer, partner-as-verb, "I'll keep this brief").
2026-05-02
"Good opener" example verdict: "yes, sounds like me."
2026-05-02

Forest's SOUL

SOUL verdict: "sounds right." Bluntness held at draft level (direct, not harsh); disagreement bar = real risk; humor = playful, not goofy.
2026-04-27
Core conviction: quality of contact trumps quantity, always.
2026-04-27

Reference library

Director Partner Network seeded — 25 directors + 28 reference films — replacing the "build a film library directly" plan.
2026-04-28

Open questions and next steps are tracked in project memory, not here — run /project-status, or see _migration/CLAUDE_CODE_MIGRATION.md for the runtime migration checklist.

BlueForest Studios Brand Films Agent
Forest — project overview