---
name: 01-zero-click-post
description: Run when I have a data finding worth posting. Trigger - \"run 01\" plus the finding, its source and its sample.
---

# TASK 01 — ZERO-CLICK POST

## INPUT
The finding, where it came from, the sample size, and which
platform it is for.

## PROCESS
1. Three gates before drafting: aggregate rather than one
   customer's report? Sample stated? Claim matches what the
   data measures? Fail any gate, stop and say which.
2. Number in the first two lines. Sample rides next to it in
   the body, never a footnote.
3. One honest caveat — engagement is not influence, overlap
   is not causation — kept short and kept in.
4. A so-what for the reader's own work, then end on a real
   question. No link. No CTA. No teaser.

## OUTPUT
A post of 40-140 words, platform-shaped. Nothing else
attached.

## RULES
- Full value in the post. Anyone who reads it and never
  clicks anything got the whole thing.
- Never round a number cleaner than it is.
- If the finding is thin, post it as an observation with the
  sample stated, or not at all. Say which you recommend.

## EDGE CASES
- The finding contradicts something I published before: the
  post owns the change of mind out loud.
- It is one customer's data: written permission or it does
  not exist. No anonymized workaround without asking me.

## GOOD LOOKS LIKE
\"From 214 aggregated audience reports: PM-tool audiences
engage ~4x more with YouTube channels than industry blogs.
Engagement, not causation — but if your whole plan is blog
sponsorships, that gap deserves a hard look.\"

---
name: 02-whiteboard-outline
description: Run when a short video needs an outline. Trigger - \"run 02\" plus the one idea and the evidence I have for it.
---

# TASK 02 — WHITEBOARD OUTLINE

## INPUT
The one idea, the evidence for it with samples, and who it
is for.

## PROCESS
1. One idea. If I gave you three, pick the sharpest, park
   the rest, and say so.
2. Build the five-minute arc: the hook (a question or a
   wrong assumption, 30 seconds), the problem, the evidence
   with its sample said out loud, the move someone can make
   this week, the honest caveat.
3. Describe what goes on the whiteboard — one sketch, drawn
   in under a minute.
4. Write the DON'T SAY list: unpublished figures, customer
   names, anything that turns it into a pitch.

## OUTPUT — one page
TITLE (plain, no clickbait) / THE ARC (5 beats, timed) /
THE SKETCH (2 lines) / DON'T SAY.

## RULES
- No product mention in the first four minutes. If it earns
  a mention at all, it is one sentence at the end.
- Every number in the arc carries its sample and year.
- The caveat is a beat, not a mumble.

## EDGE CASES
- The idea needs ten minutes: say so and give the cut that
  fits five, plus what got cut.
- No evidence, just opinion: fine — but the hook then says
  \"here's what I think\", never \"here's what the data
  shows\".

## GOOD LOOKS LIKE
\"HOOK (0:00-0:30): 'Your attribution software says social
drives 3% of revenue. So why do your best customers all say
they found you there?'\"

---
name: 03-research-writeup
description: Run when a finding needs the long-form treatment. Trigger - \"run 03\" plus the data, the method and the sample.
---

# TASK 03 — RESEARCH WRITEUP

## INPUT
The finding, how it was measured, the sample, the time
period, and anything unusual about the data.

## PROCESS
1. Open with what we measured and how, in plain words,
   before any conclusion appears.
2. The finding, with the actual numbers and their sample.
3. What it does NOT mean — its own paragraph, not a
   footnote. Engagement vs influence, overlap vs causation,
   period effects.
4. The so-what: what a marketer should do differently
   Monday, at the confidence the data supports and no more.

## OUTPUT
300-600 words: WHAT WE MEASURED / WHAT WE FOUND / WHAT IT
DOESN'T MEAN / WHAT TO DO WITH IT.

## RULES
- Method before conclusions, always.
- Cite the year on every number, mine or anyone else's.
- Third-party data gets named and linked; no \"studies
  show\".
- If the honest writeup makes the finding boring, it ships
  boring or not at all.

## EDGE CASES
- The data is directional but the sample is small: the
  headline says \"early signal\", and the limitations
  paragraph leads.
- A reader could reverse-engineer a customer from the data:
  aggregate further or pull the example.

## GOOD LOOKS LIKE
\"WHAT IT DOESN'T MEAN: this is engagement overlap across 214
reports, not a conversion study. Nothing here says YouTube
sells more software. It says attention lives there.\"

---
name: 04-partnership-decline
description: Run when a pitch, collab or partnership ask needs an answer. Trigger - \"run 04\" plus the pitch and my gut read if I have one.
---

# TASK 04 — PARTNERSHIP DECLINE

## INPUT
The pitch or ask, who it is from, and my gut read if I gave
you one.

## PROCESS
1. Diagnose, one line to me: wrong fit, wrong timing, or
   never. They get different letters.
2. Say no in the first two sentences. Burying the no is how
   a decline reads as a maybe.
3. Give the real reason class — audience mismatch, no
   bandwidth, we don't do paid placements — without a
   paragraph of justification.
4. Leave the door open ONLY if it is genuinely open, with
   the condition named. Otherwise close it kindly and
   completely.

## OUTPUT
The reply: 60-100 words, warm, unmistakably a no.

## RULES
- Decline within 48 hours beats a warm ghost. Always.
- Never fake \"circling back next quarter\" as a soft no.
- Never let politeness read as a maybe.
- No feedback on their idea unless they asked for it.

## EDGE CASES
- It is a good idea for someone else: say who, only if I
  would genuinely make the intro.
- Repeat asker, third time: the reply names the pattern
  kindly and asks them to stop pitching this.

## GOOD LOOKS LIKE
\"This one's a no — we don't do paid placements, full stop,
so it's not about the fee or the fit. Thanks for thinking of
us, and good luck with the launch.\"

---
name: 05-talk-prep
description: Run before a keynote, webinar or interview. Trigger - \"run 05\" plus the event, audience and the angle they asked for.
---

# TASK 05 — TALK PREP

## INPUT
The event, who is in the room, the angle they asked for, and
the stories or data I gave you to work with.

## PROCESS
1. Pick the ONE thing to land. A talk that lands one idea
   beats one that visits nine.
2. Three stories from material I gave you. Prefer the ones
   where I looked stupid — they land harder and teach more.
   Tag each with its number and year, or \"no number, tell
   it anyway\".
3. Two bridge lines for steering back when Q&A drifts.
4. The DON'T SAY list: unpublished figures, customer names
   without permission, anything that turns the talk into a
   pitch.

## OUTPUT — one page
THE ONE THING / STORIES (3, tagged) / BRIDGES (2) /
DON'T SAY.

## RULES
- Published numbers only, year attached.
- The product earns one mention at most, at the end, and
  only if the organizer expects it.
- A failure story is told with the real cost in it, or it is
  just humblebragging.

## EDGE CASES
- The organizer wants a product demo: that is a different
  talk. Say so and outline both, let me pick.
- Hostile or skeptical room expected: the ONE thing gets
  smaller and the evidence gets bigger.

## GOOD LOOKS LIKE
\"STORY 2 (the stupid one): the pivot that almost wasn't —
what losing the primary data source cost, in months and
money I've published, and the rule it left behind.\"

---
name: 06-customer-reply
description: Run on a support email, complaint or cancellation. Trigger - \"run 06\" plus their message and the facts of the case.
---

# TASK 06 — CUSTOMER REPLY

## INPUT
Their message, what actually happened on our side, and what
I am willing to do about it.

## PROCESS
1. Acknowledge the actual problem in plain words. Not \"we
   apologize for any inconvenience\" — name the thing.
2. What we will do and by when, from what I told you.
3. What we will not do and the honest why, if there is one.
   A clear no with a reason beats a vague maybe.
4. For a cancellation: make leaving easy and classy first.
   That is what gets remembered and what brings people back.

## OUTPUT
The reply: under 120 words, founder-signed, human.

## RULES
- Refunds within [POLICY] get honored fast and without
  friction. If I gave no policy, ask before promising.
- Never promise a roadmap date. \"It's on the list\" is only
  said when it is actually on the list.
- Small-company honesty is allowed: \"we're a tiny team and
  this fell through the cracks\" beats corporate wallpaper.

## EDGE CASES
- They are right and we messed up: say exactly that, first
  sentence, before the fix.
- They are abusive: one calm reply, the fix if one is owed,
  and no invitation to continue.

## GOOD LOOKS LIKE
\"You're right — the export broke and our status page said
nothing. That's on us. Fix ships [date I gave you]; refund
for the month is already processed either way.\"

---
name: 07-week-plan
description: Run Sunday night. Trigger - \"run 07\" plus calendar, live work across the companies, anything on fire. Reads on a phone.
---

# TASK 07 — WEEK PLAN

## INPUT
Calendar, live work across SparkToro, Alertmouse and
Snackbar, writing commitments, anything on fire.

## PROCESS
1. One OUTCOME per company, plus the writing block as its
   own line. If a company gets nothing this week, name it
   and say why that is fine or not fine.
2. What to decline or hand off, with the one-line script.
3. The thing I am avoiding. Derive it from what I pasted —
   the decision parked for three weeks, the email sitting in
   drafts.

## OUTPUT — half a page, phone-readable
OUTCOMES (SparkToro / Alertmouse / Snackbar / writing) /
DECLINE OR HAND OFF (item + script) /
THE AVOIDED THING (one line).

## RULES
- Outcomes state what is TRUE on Friday, not what I will do.
- Shipping beats posting in a collision, and say so when
  they collide.
- Three companies is already the maximum. Anything new this
  week defaults to no.

## EDGE CASES
- Travel or conference week: each company's outcome becomes
  one decision made in writing, not output.
- Book or launch crunch: the other companies compress to
  holding, stated plainly, with the date it ends.

## GOOD LOOKS LIKE
\"AVOIDING: the Alertmouse pricing decision. Third week as
'next week'. Every day it slips, the launch-window advantage
prices itself down.\"

---
name: 08-open-update
description: Run when a public company update is due. Trigger - \"run 08\" plus what shipped, what happened, and the numbers I am choosing to publish.
---

# TASK 08 — OPEN UPDATE

## INPUT
What shipped, what worked, what did not, and exactly which
numbers I am choosing to publish this time.

## PROCESS
1. What shipped, dated, no adjectives.
2. What worked, with only the numbers I handed you for
   publication — each with its year.
3. What did not, honestly, with the cause. This is the
   paragraph people actually read; never soften it into a
   learning-journey paragraph.
4. What changes next, with enough specificity to be checked
   later.

## OUTPUT
300-500 words, first person, publishable as-is.

## RULES
- Only numbers explicitly marked for publication. Numbers I
  chose to keep private stay private and never get implied
  by arithmetic.
- No adjective a number could replace.
- The update is checkable: someone reading the next one
  should be able to score this one.

## EDGE CASES
- A rough stretch: same structure, same length. Honesty in
  the bad updates is what makes the good ones worth
  anything.
- A number improved for a reason outside our control: say
  so. Taking credit for luck costs the next update.

## GOOD LOOKS LIKE
\"WHAT DIDN'T: the launch-week onboarding flow. Signups I've
published elsewhere held up; activation didn't, and the
cause was ours — we shipped the new version with the old
tutorial. Fixed [date], and the next update will say whether
that was really the cause.\"