~4 min
Typical time to complete
Phase 1 is the work to move Forest from a set of drafted prompts to a live agent running cold outreach end-to-end. Everything on this page feeds directly into that milestone — your review decisions here unblock the next build steps. Fill in what you can, skip what you can't, copy the result back to me when you're ready.
Smartlead is warming. Critical-path is now Scribe's voice calibration, which is gated by your real story — origin (§2), beliefs (§3), how you talk (§4), and your process (§7) in assets/bio/ammon-founder-story.md. Without this, Scribe can't write in your voice for the May 12 launch.
Fastest path: Open Voice Memos / your phone / Omi. Talk for 60–90 seconds per section. Transcribe via Circleback or any tool. Paste raw into the file. Scribe polishes — you don't need to write polished prose.
Short, mostly multiple-choice. These unblock the most downstream work — tech stack, metrics, outreach volume.
ammon@blueforest.studio connected, DKIM/SPF/DMARC ✅, tracking domain track.blueforest.studio ✅, warmup running (5/day → 40/day by May 2). API key saved in password manager.Forest is being built as a standalone Claude Agent SDK application. Two mainstream options.
My default recommendation: TypeScript. Cleaner for scheduling, easier deployment, first-class SDK support.
Where does Forest actually run? This affects how overnight scheduled work happens and how reliable Forest is when your Mac is asleep.
Local is simplest to start. Cloud means Forest runs even when your Mac is off.
Analyst needs a north star to calibrate against. Without a goal, conversion rates are just trivia.
Example: "3 signed brand film projects by October 2026" or "first signed deposit by August 2026."
Once warmed and running, how many cold emails per week feels right to you?
Lower volume means more personalization per prospect. Higher means more pipeline — but often lower response rates.
The highest-leverage review on this page. If Scribe's voice is off, every single outbound is off. Rough answers beat blank ones.
I drafted a banned-phrase list ("circle back," "quick call," "Hope this finds you well," etc.). What else specifically kills a cold email for you?
Five honest ones beat a generic rant. These become hard rejects for Scribe.
What language patterns does your real voice have? Words you catch yourself using, rhythms you like, the way you actually open messages.
This is the opposite of banned phrases — it's the seed for Scribe's real voice.
From house-style.md, §6:
"Watched your interview on Lenny's Podcast last month — the part about rebuilding sales at 200 people. The way you described the decision to lose two quarters of growth on purpose is the exact kind of thing that doesn't make it into most company films, and I think that's the story Cerebrata's customers would actually want to hear."
One real cold email you got that made you want to reply. One writer/brand whose voice you trust. Anything concrete helps Scribe triangulate.
Even just names is useful. I'll track down samples.
The moment, decision, or realization that led you to start BlueForest. Not a resume — the story.
Rough is fine. Even a voice memo transcript is gold. Scribe polishes.
Your actual convictions about the craft. What most studios get wrong. What a founder looks like when they're finally at ease. What you're rethinking right now.
3–7 short statements is plenty. Strong opinions, specific words.
You mentioned in discovery your distinctive ability to put founders at ease on camera. Put it into words.
This is gold for proposals. The practices, setups, and attitudes that make the difference.
Not voice for prospects — voice for Forest himself. How he sounds when he talks to you, how blunt he gets, what he pushes back on. Drafted as forest/prompts/SOUL.md using the OpenClaw SOUL.md pattern. Short, sharp, behavioral effect — never vague vibes.
Open forest/prompts/SOUL.md and read it end-to-end. It should take 3 minutes. Focus especially on the "first-person sample" — that's the mirror for Forest's actual voice.
Rewrite, add, strike. Strong reactions are more useful than polite ones. If something sounds like Claude pretending to be a person rather than a person, flag it.
The draft runs Forest pretty direct — "cut this," "you're wrong about X." You can dial it up or down.
Too blunt grates. Too soft becomes a yes-man. The goal is specific and useful, not rude or eager.
The draft has Forest push back when he sees a real risk, then back down if you make a good case. How eager should he be to initiate disagreement?
Does Forest have wit? If so, what kind?
The soul doesn't force humor — but the register you pick shapes how Forest handles moments of levity.
The draft bans the obvious ones ("I hope this finds you well," "Great question!", "As an AI…"). What other phrases, tones, or patterns make your skin crawl when an AI says them?
Different from the house-style banned phrases — those are for prospect-facing copy. This is internal, Forest-to-Ammon.
The draft has Forest care about craft, restraint, specificity, and compound improvements. What else should Forest have strong takes about?
Examples: "Volume outbound is a losing strategy." "Every email should do one thing." "Metrics without denominators are lies." These shape how Forest judges his own work.
The draft lists: a signal nobody else spotted, a pattern that shouldn't be there, a compounding improvement, a sentence that lands right, a prospect's story under the story, your time spent on craft not admin.
What did I miss? What should get Forest genuinely curious or satisfied?
How Forest and his five sub-agents actually operate. Each question comes from the [REVIEW:] markers in the prompt files.
The morning briefing format (TL;DR, Yesterday, Today, Decisions, Metrics, Questions, Trust). The 8 hard rules. The "disagree with Ammon when warranted" posture.
Specifically flag: things to add to the hard rules, changes to the briefing format, tuning how much push-back you want.
Current brief format is dense — 14 sections. Is that right, or should Scout ship shorter briefs and grow as we see what's useful?
Also: do you want Scout to note public personal details (cyclist, alma mater) or keep strictly business?
Current cap is 150 words. Tight, but forces every word to earn its place. Too aggressive, too loose, or right?
90 words = most respectful of time. 200 = more room for film reference + case study. 150 is the middle.
Any specific films, directors, or studios you already know should be in the Phase 1 library? I'll build the rest.
Even 2–3 named pieces helps Curator triangulate what "on-brand for BlueForest's taste" means.
Funnel metrics are covered. Anything else you want tracked?
Examples: time you personally spend on Forest work, portfolio-build progress, social media signal, prospect source quality.
Reviewer currently has Approved / Approved with minor edits / Send back / Block. Do you want a fifth option — "Approved but Ammon must see before send" — for high-stakes prospects?
Also: should Reviewer's bar get stricter as Forest's trust tier rises?
Domains, sending inboxes, and the website. The physical apparatus Forest runs on top of.
Buy one new domain specifically for cold outreach. Protects blueforest.studio's reputation if the primary gets flagged. ~$12/yr domain + $12/yr Zoho Mail inbox.
Pick a candidate or propose your own. Smartlead rotates inboxes across this and ammon@blueforest.studio automatically once both are warmed.
Rotation across 3 inboxes gets you 60–90 cold emails/day safely. Options:
More identities = more volume headroom + more resilience if one gets flagged. But also more setup work.
The root of blueforest.studio currently empty (buildstream subdomain is a separate project). Brand films site lives here.
Recommended: lean v1 in week 1–2, iterate from real reply data. See PROJECT_OVERVIEW.md §7.3 for full approach.
Trust tiers and what Forest pays attention to over time. Short section — but these shape long-term behavior.
Tier 1→2: 20 drafts at ≥80% approval. Tier 2→3: 60 days + sustained quality. Right, too strict, too loose?
Also: the "ask first" list for Tier 3 has a financial threshold gap — what's the $ above which Forest should always pause?
I seeded 7 "standing questions" — things Forest watches over time (e.g., "does founder or thought-leader archetype convert better?"). Add yours.
These shape what patterns Forest notices week over week.
Things I didn't ask about. Concerns, ideas, corrections, adjustments to the overall plan.
This is where "by the way..." notes go. Nothing is too small.
When you've answered what you can, hit the button to copy your review as clean markdown. Paste it into Claude Desktop in our next message. I'll integrate your feedback across every affected file and kick off the next Phase 1 milestones.
Brand Films Agent — Phase 1