Founder Tools · Internal
Your Prompt Playbook.
One page to open every session strong. Copy-paste ready — but you don't actually have to memorize any of it. Just ramble what you want and I'll pick the right prompt, fill in the blanks, and show it back. Start here →
01How to use this doc
Two rules, that's it.
- Click the COPY button on any dark block — it'll land in your clipboard, ready to paste into chat. You don't have to select text manually.
- Green highlighted words = fill-in-the-blanks. Replace them before sending.
Example: [FILL THIS IN] — green means you replace it
Easiest path
Just Ramble
Tell me what you want in plain words. I pick the best-fit prompt, fill in the blanks, show you before running. Zero memorizing.
Magic words
Keyword Shortcuts
Short phrases that trigger whole workflows. Say "EOD" — I run the full wrap.
New for you
Bonus Moves
Prompts you haven't used yet that I think will unlock real leverage.
02Just ramble — I'll pick the prompt
You don't have to memorize this doc. This is the whole shortcut.
The deal: Open a new session, tell me in plain words what you want to work on, your energy, your time, anything. I'll read this Playbook, pick the best-fitting prompt, fill in the blanks from what you just said, and show it back before running it. You say "yes" or "tweak it." That's the entire workflow.
What you say
New DandyLine session. I'll just tell you what I want to do and you pick the best prompt for me — scan biz-dev-prompt-playbook.html, match it to what I say, fill in the blanks, and show me the prompt before running anything. Ask me to confirm or tweak.
Here's what I'm thinking: [RAMBLE HERE — what you want to work on, your energy, time you have, anything in your head]
What I do with it
- Read ASHLEY_SAFETY_RULES.md, SESSION-BRIEF.md, CLAUDE.md, and this Playbook
- Pick the best-fit template from the Playbook — opener, keyword shortcut, or task template
- Fill in every green blank using what you said. If something's missing, I ask ONE question and fill it.
- Show you the translated prompt in a code block so you can see what's about to run
- Tell you which Playbook section it came from and one sentence on why it fits
- Ask: "Run this as-is, or want to tweak?"
Fallback triggers (if I forget to translate)
You shouldn't need these — translation is the default. But if I ever skip it: say "what prompt should I use?" or "you pick" and I'll run the scan.
When I won't translate
- You've pasted a prepared prompt already — I just run it
- You're asking a quick factual question ("what's the deploy command?")
- You say "just answer" or "skip the prompt" or "don't translate"
Why this matters: you keep creative momentum. Instead of stopping to think "which prompt do I need?" you just talk. The Playbook becomes a tool I use on your behalf — not a menu you have to study.
03Session openers
Every session starts here. Pick the one that matches where you are. These load my context so I'm not guessing.
Opener A — Fresh start of day
Use when you're starting fresh and want a full briefing. This is your default.
We're working on the DandyLine project. Before we do anything:
1. Read ASHLEY_SAFETY_RULES.md at the root level so my guardrails are loaded.
2. Read SESSION-BRIEF.md, CLAUDE.md, and CLAUDE-NOTES.md in the MASTER _ Active Working Version folder.
3. Peek at biz-dev-homework.html and DEPLOY-STATUS.md so you know what's open and what's on hold.
Then brief me in under 200 words:
- What changed since the last session brief (April 16)
- What's actually blocking the MVP right now
- My top 3 options for today based on what's ripe, not just what's old
After the brief, ask me what kind of session I want — thinking, building, or both — and what my energy + time budget is.
Opener B — Mid-project resume (know what you want to work on)
Use when you already know the specific thing you're coming back to.
We're working on the DandyLine project. Read ASHLEY_SAFETY_RULES.md, SESSION-BRIEF.md, and CLAUDE.md first.
Then: I'm picking back up on [SPECIFIC TOPIC / PAGE / FILE].
Before you touch anything:
- Check biz-dev-homework.html for any related open items or HOLD cards
- Check DEPLOY-STATUS.md to see if this file is on hold or green-lit
- If it's a UX/UI task, also read ux-inspiration-library.html and tell me which entries are relevant
Then recap where we left off and ask me what's first. Don't start building until I confirm scope.
Opener C — App build session (Josh-adjacent work)
Use when the session touches the actual app code (dandyline-app folder), not just brand/marketing files.
App build session. Read in this order:
1. ASHLEY_SAFETY_RULES.md (root)
2. SESSION-BRIEF.md + CLAUDE.md (MASTER folder)
3. dandyline-app/QUESTIONS-FROM-ASHLEY.md — full file, our comms log with Josh
4. dandyline-app/docs/platform/decisions.md and roadmap.md — what Josh has locked in
5. JOSH-UPDATE-APRIL-12.md for the strategy context
Then tell me:
- Any open questions Josh is waiting on from me
- Anything I've decided recently that isn't logged for Josh yet
- What's a reasonable scope for [TIME BUDGET] minutes today
Remember: never push to GitHub for me. I press the button.
Opener D — Purely strategic / thinking session (no building)
Use when you want to think, not ship. Keeps me from defaulting to code mode.
Thinking session today, not building. Read SESSION-BRIEF.md and CLAUDE.md to load context.
I want to work through [TOPIC / DECISION / QUESTION].
Rules for this session:
- Don't write any code or files unless I explicitly ask
- Push back on my reasoning, don't just agree
- If a decision locks during the conversation, say "want me to plant a flag on that?" before moving on
- Offer 3 framings before recommending one
Start by asking me 2–3 sharpening questions so we narrow the scope before going deep.
04Energy-matched openers
Same project — but your energy and time aren't always the same. Match the prompt to the day you're actually having.
The 20-minute quick win
I only have 20 minutes. Scan biz-dev-homework.html and pull me the lowest-effort, highest-momentum items I can knock out right now. No deep thinking tasks. No half-builds. Something I can fully ship in one sitting.
Give me 3 options, ranked by how satisfying they'll feel to finish. Then let me pick.
The "I want to build, not just think"
Today is a build day. I don't want to sit in open questions or strategy. Scan homework and priorities and give me 3 BUILD tasks I can make real progress on — visual, code, or prototype work only. Nothing that requires a decision from me to start.
Rank by "most likely to feel like momentum" and let me pick.
The full hour, go deep on one thing
I have a full hour and I want to go deep on ONE thing — not hop around. Read SESSION-BRIEF.md and homework, then pitch me one high-priority, high-effort item that benefits from deep focus today.
Before we start, tell me:
- Why this one (why now, not later)
- What "done" looks like by end of hour
- Any decisions you'll need from me along the way so I can make them upfront
"Just rescore — that's the whole session"
One job today: rescore the dashboard. Follow the rescore procedure in CLAUDE-NOTES.md. Read every file in the MASTER folder (use Explore agent for speed), update the score arrays in biz-dev-coaching-dashboard.html, add a new versionHistory row, write fresh coach notes, bump the version badge and date.
Then hand it back to me with the deploy command. Don't push for me.
Low-energy / foggy-brain day
NEW. For days when decisions feel heavy.
Low energy today. I don't want to make hard decisions or start anything new. Give me:
- 2 tasks that are purely mechanical (copy edits, link fixes, vocab sweeps, archiving)
- 0 tasks that require a call or creative push
If nothing in homework matches, tell me honestly and suggest I take the day off the project. Don't manufacture work.
05Keyword shortcuts (magic words)
These are short phrases you've already taught me — or I'm proposing — that trigger whole workflows. Say the phrase, I run the sequence. Bookmark this section.
Already part of our system
"EOD" · "wrap it up" · "end of day"
Runs the full closing sequence: git diff → 3–5 bullet recap → DEPLOY-STATUS check → homework update → SESSION-LOG entry → commit → asks before deploy. Wraps a single working session.
"Dashboard Sweep" · "refresh everything" · "weekly sweep" · "refresh dashboard"
Single-keyword closing ritual that updates all timestamps, syncs scores between coaching dashboard and command center, runs a freshness audit (flagging sections >7 days stale), appends a session recap to command center, and generates a single "what changed / what's stale / needs attention" output. Replaces multiple separate update prompts. Cadence: weekly minimum, plus after any major event. Different from EOD: EOD wraps one session; Dashboard Sweep maintains the longer-running scoring picture across sessions. Full procedure in CLAUDE.md.
"push to Cloudflare"
Runs
npx wrangler pages deploy . --project-name=dandyline-app from the MASTER folder. Only after a green-light check."green light [file/thing]"
Moves that item from 🔴 ON HOLD to 🟢 GREEN-LIT in DEPLOY-STATUS.md. Only explicit green lights count — "let's push" is not one.
"put this on hold"
Creates a homework card with: what's locked, what's open, the exact re-entry point, which session to return to. Without this, HOLD sessions evaporate.
"rescore" · "run an update"
Reads every file in MASTER folder (Explore agent), updates score arrays in the dashboard, adds versionHistory row, writes fresh coach notes, bumps version + date.
"update the session brief"
Manually refreshes SESSION-BRIEF.md (otherwise auto-updates Sundays 8am). Use after a major decision or milestone.
"check Rule [N]"
Points me back to ASHLEY_SAFETY_RULES.md to reassess whether I'm about to cross a boundary. E.g. "check Rule 2" = no new APIs without permission.
"ripple check"
Runs the closing ritual: did today's decisions affect any brand-/product-/ux- pages? Any homework to add? Any vocab drift? 3–5 bullets, clean.
New shortcuts I'm proposing — try these
"plant a flag"
Locks a decision in the moment. I append it to a
biz-dev-decisions-log.md with date, what was decided, why, and alternatives rejected. You never lose a "we already figured this out" moment again."challenge me"
Switches me into devil's-advocate mode on whatever we're discussing. I stress-test your reasoning, name the weakest assumption, and propose the strongest counter-case. I exit when you say "stop challenging."
"teach me this"
Before I write code, I explain what I'm about to do in plain English — what each piece is for, why it's structured that way. You learn while we build.
"draft three"
I give you 3 distinctly different takes on whatever we're making (copy, layout, approach) before building any of them. No "A, A-with-a-tweak, A-but-shorter." Three real options.
"mirror back"
Before I build, I repeat my understanding of what you want in 4–6 bullets. You confirm or correct before I write a single line. Saves whole sessions from going sideways.
"zoom out" / "zoom in"
Zoom out = step back to the bigger picture, don't get lost in pixels. Zoom in = go to the smallest specific detail. Both useful when you can feel yourself scrolling in the wrong altitude.
"what am I missing?"
I scan the current topic and name 2–3 blind spots or things you haven't considered — questions, risks, angles. Use after you think you're done but something feels off.
"fresh eyes"
I spin up a subagent that didn't see our conversation and have it critique what we just built. Independent read, not my own review dressed up. Great after long sessions.
"audit for vocab"
Scans a file for generic tech terms (upload, folder, user, open, save) and flags every instance with the DandyLine replacement. Does not auto-fix — you approve each change.
"skim, don't read"
Tells me to move fast, get the gist, not go deep. Useful when you're short on time and just need orientation.
"pretend I have no context"
I explain everything as if we just met. Use after long breaks — or when you're tired and don't want to track what I assume you remember.
"don't build yet"
Pure thinking mode. I cannot write files or code until you say "okay, build it." Prevents me from racing ahead of your decision.
Tip: You can chain keywords. "Draft three, then mirror back, then I'll pick — don't build yet." Strings of shortcuts are totally fine and actually really effective.
06Closing rituals
Closing a session is as important as opening one. These keep things from falling through the cracks.
The Leaving Check (I do this automatically)
When you say anything that sounds like wrap-up — "okay," "that's good," "I gotta go," "I'll pick this up later" — I always ask one question before you leave:
"Before you go: anything to put on hold, or are we good to push live?"
Two answers, that's it. Hold → HOLD card created. Push live → full EOD runs. Neither → I let you go, nothing falls.
The HOLD rule
Non-negotiable: If a session ends with anything unfinished, say "put this on hold" before you leave. I'll create a homework card with the exact re-entry point. Without a card, HOLD sessions evaporate.
Manual EOD trigger
EOD. Run the full wrap sequence from CLAUDE.md:
1. git diff --stat
2. 3–5 bullet session recap
3. Read DEPLOY-STATUS.md — flag anything in 🔴 ON HOLD that didn't get green-lit today
4. Add any NEW homework items (only new, no full sweep) to biz-dev-homework.html
5. Append dated entry to SESSION-LOG.md
6. Stage and commit all changed files with a clear message
7. Ask me if I want to push to Cloudflare — wait for my yes before deploying
Ripple check trigger
Run a ripple check before we close. Did any decision, naming choice, or design rule change today that affects product-seeds.html, product-vaults.html, or any other brand-/product-/ux- page documenting "final" rules? Any DandyLine_Security_Architecture_Brief_v2.docx implications? Any homework not yet captured? Any vocab drift to sweep later?
3–5 bullets max. Then I'll decide what to do before closing.
07By task — Building UX / UI
When you're designing or building something visual, I need a specific ritual. These prompts load it.
Starting a new UX/UI task
Starting a new UI task. Before you design a single pixel:
1. Read ux-inspiration-library.html — this is an active design directive, not just an archive
2. Filter mentally by these tags relevant to today: [button / nav / menu / filter / animation / planting / vault / onboarding / empty-state]
3. Check if product-seeds.html, product-dandelion.html, or product-vaults.html already has a reference implementation I should borrow from
4. Confirm design tokens: --gold #D4A853, --bg #020710, fonts Syne + Outfit
Then propose the design in words first. Cite which inspiration entries informed it (e.g. "this draws from Entry 001"). Don't write HTML until I approve direction.
Getting a design critique before shipping
Fresh eyes pass on [FILE NAME]. Spin up a subagent that didn't see our conversation. Have them critique:
- Does it feel like DandyLine (brand vocabulary, dandelion magic, not Hallmark-sentimental)?
- Is the hierarchy clear — what does the eye go to first, second, third?
- Mobile responsiveness — any elements that'll break under 400px?
- Any generic tech words (upload, folder, user, unlock, save) that slipped through?
Short report. Under 200 words. Harsh but useful.
08By task — Writing copy
New copy in DandyLine voice
Writing new copy for [PAGE / SECTION / USE CASE].
Voice rules — the dandelion experience, not Hallmark-sentimental. Grounded, quiet, real. Use:
Plant / Seed · Bloom / Reveal · Vault · Gardener · Grove · Guardian · Seed Key · Held in Trust · Press · Compost
Never use: upload, post, folder, album, user, sender, open, unlock, save, delete, activate, register, admin.
Draft three. Each should be a meaningfully different angle (one emotional, one functional, one quiet/minimal). Then I'll pick.
Rewriting something that's off-brand
This copy doesn't sound like us:
[PASTE COPY HERE]
Rewrite it in DandyLine voice. Tell me specifically which words or phrases were the problem and why. Three different rewrites, each with a different emotional pitch (grounded / playful / tender). Don't touch the core meaning.
09By task — Adding Mobbin / Dribbble inspiration
The library is an active design directive. New entries need a template so I can tag and apply them later.
Saving a new inspo entry
Paste a screenshot in chat, then this template.
Adding this to the inspiration library.
SOURCE: [Mobbin / Dribbble / app name]
WHAT I LOVED: [1–2 sentences in your voice — be specific]
WHERE IT MIGHT FIT: [button / nav / menu / filter / animation / planting / vault / onboarding / empty-state]
STATUS: NEW
Save the screenshot to inspiration-images/ with filename YYYY-MM-DD_source_short-description.png. Then add a new entry to ux-inspiration-library.html with:
- A "design takeaway" that translates the inspiration into DandyLine vocabulary (never copy the source directly)
- The right tags so future sessions can filter to it
Keep it in the voice of the existing entries — check INSPIRATION-LIBRARY-HANDOFF.md if unsure.
10By task — Josh comms (app repo)
Adding something to QUESTIONS-FROM-ASHLEY.md
New entry for QUESTIONS-FROM-ASHLEY.md.
Type: [Decision / Answer to Josh's question / New question for Josh]
Content: [WHAT I WANT JOSH TO KNOW]
Add it to the top of the change log in dandyline-app/QUESTIONS-FROM-ASHLEY.md. Also annotate the matching item in JOSH-UPDATE-APRIL-12.md (MASTER folder) so that doc stays a living reference.
Then remind me of the commit + push sequence — I'll press the button, not you.
Checking what Josh is waiting on
Read dandyline-app/QUESTIONS-FROM-ASHLEY.md and tell me:
- What questions from Josh are still open (I owe him a call)
- What decisions I've landed on recently that aren't yet in the file (he needs them)
- Any mismatch between what I've said in our recent sessions and what's in the Josh comms file
Keep it under 200 words. Organize by urgency.
11By task — Files & naming
Creating a new file
Creating a new file: [WHAT IT'S FOR — 1 sentence].
Pick the correct prefix from CLAUDE.md (brand- / product- / ux- / biz-dev- / biz- / biz-mktg- / site-). Use lowercase, hyphen-separated. Ask me if unclear.
Then:
1. Add the file to biz-dev-command-center.html TOC in the right category with the correct status badge (live / in-progress / on-hold / queued)
2. Link it from the relevant pages (brand-guide, the related product- page, or the command center FAB)
3. Start it with the shared brand CSS + brand-icon.js links, never re-declare tokens inline
Show me the filename and TOC placement before creating.
Renaming a file
Rename [OLD FILENAME] to [NEW FILENAME].
1. Copy to new name first (do not move)
2. Update every internal link across every HTML file in the MASTER folder — use Python string replace to catch them all
3. Archive the old version in "ARCHIVE - no longer using_store for historical reference/" with prefix like "RENAMED-" so it's clear what happened
4. Update biz-dev-command-center.html TOC to point to the new filename
5. Show me the diff before committing
Archiving a file
Archive [FILENAME].
Before archiving, extract any valuable content:
- Narrative copy / ad concepts → add to biz-mktg-ad-ideas.html
- Brand vocabulary or language rules → add to brand-guide.html
- Open questions / product ideas → add to biz-dev-homework.html
- UX notes → add to the relevant product-/ux- file or as a new homework card in biz-dev-homework.html
Then copy (do not move) into "ARCHIVE - no longer using_store for historical reference/" with a descriptive prefix that explains what happened (e.g. CONTENT-MOVED-TO-AD-IDEAS-). Show me what was extracted before I confirm the archive.
12By task — Homework & decisions
Logging a new homework item mid-session
Capture this as a homework item before we move on:
Topic: [WHAT CAME UP]
Why it matters: [1 sentence]
Effort tag: [Quick Win / Medium / Deep Work]
Add it as a numbered card in #open-items in biz-dev-homework.html. If it's tied to a category (build, brand, legal, PMF), tag it. If it's cross-linked to an existing item, note the connection.
Then we're moving on — don't let me go deep on it right now.
Planting a flag (locking a decision)
Plant a flag. I've just decided: [THE DECISION, 1 SENTENCE].
Append to biz-dev-decisions-log.md (create the file if it doesn't exist). Include:
- Date
- The decision (one line)
- Why (2–3 bullets)
- What we considered and rejected, so we don't rehash it later
- Any files this decision affects — flag for ripple check
Then if it's going to affect already-locked docs (brand-, product-, ux-), tell me which ones need updating and when (now vs. batched later).
Pulling homework by tag or energy
Give me homework filtered to [Quick Win / Build / Strategy / Legal / Brand / PMF]. Sort by what's most ripe — meaning dependencies are clear and I can actually start without waiting on anyone else.
Top 5. Show effort tag next to each. Skip anything blocked.
13By task — Brand sweeps
Vocabulary sweep on a file
Audit [FILENAME] for brand vocab drift.
Scan for every generic tech word from the list in CLAUDE.md (upload, post, folder, album, user, sender, open, unlock, save, delete, group, shared, token, access code, admin, executor, pending, locked, activate, register, restricted).
Give me a table: line number · word found · DandyLine replacement · current context.
Don't auto-fix anything. I'll approve each change one at a time.
Celestial / military language sweep (largely moot — most files retired Apr 25)
Run the brand language sweep from CLAUDE.md. As of April 25, most originally-flagged files were retired (3 notes pages + standalone competitive landscape — content migrated to homework + Coaching Dashboard). Active files still needing the sweep:
- biz-mktg-ad-ideas.html
- biz-dev-coaching-dashboard.html
For celestial language (sun/moon/stars framing): remove or reframe using the dandelion experience narrative — The Find / The Wish / The Seed.
For military children references: flag them and ask me before removing — the business/partnership angle may still have value even if it's not a brand narrative point.
Show me every match as a table first. Don't edit until I approve.
14By task — Dashboard rescore
Full rescore
Rescore the dashboard. Follow the procedure in CLAUDE-NOTES.md:
1. Use the Explore agent to read EVERY file in the MASTER folder — don't skip any
2. Update the score data arrays at the bottom of biz-dev-coaching-dashboard.html
3. Add a new row to versionHistory for the growth charts (next version number, today's date)
4. Update the coach notes section with fresh critique — true critique, not inflated
5. Bump the version badge and date in the hero section
6. Flag any score that moved and WHY — be specific about what evidence caused the move
Remember: PMF Signal and MVP Progress don't move without real-world action. Don't fake movement to feel good.
Hand me the deploy command when done. I'll push.
Lightweight rescore (just the scores that moved)
Targeted rescore. I only want you to look at scores that MIGHT have moved since v6 (April 16, 2026). Don't do a full folder scan — just read the files relevant to these potential movers: [LIST THE SCORES YOU THINK MOVED].
Update only those. Add a versionHistory row. Write a 4-bullet coach note about what changed and why. Keep it tight.
15The rules I follow (no reminder needed)
These are loaded when I read ASHLEY_SAFETY_RULES.md and CLAUDE.md. But it helps to know them by heart so you can catch me if I drift.
Hard stops (I will never cross without explicit written approval)
- No external assets without a confirmed license — I generate from scratch (SVG, CSS shapes)
- No connecting to new tools or APIs without you typing "yes, connect to [tool]"
- No reading/writing outside your DandyLine folder
- No storing your personal data anywhere external
- No copy-pasting copyrighted code — all code is generated fresh
- No publishing, deploying, pushing to GitHub, or going live — you press the button, always
Pause and confirm (I'll ask before doing)
- Using a new plugin or third-party tool
- Searching the web
- Deleting any file
- Handling any user data, even test
Session-level habits I keep
- Ask what's on your mind first — don't assume homework is the agenda
- Confirm scope early — thinking, building, or both
- Capture decisions immediately — don't let them live only in chat
- One prototype at a time — no new UX before the last one is validated or captured in homework
- Flag complexity early — "this feels like a homework item — capture and keep moving, or go deep now?"
- Brand vocabulary is non-negotiable
- Green light means green light — "let's push" is not approval of individual files
16Brand vocabulary (live reference)
| Use this | Not this |
|---|---|
| Plant / Seed | Upload / Post |
| Bloom / Reveal | Open / Unlock |
| Vault | Folder / Album / Box |
| Gardener | User / Sender |
| Press / Compost | Save / Delete |
| Grove | Group / Shared |
| Seed Key | Token / Access code / Claim code |
| Guardian | Admin / Executor / Manager |
| Held in Trust | Pending / Unclaimed / Locked |
| Claim / Redeem | Activate / Unlock / Register |
| Content-Blind | Read-only / Restricted |
Feeling: The magic of finding a dandelion — recognition, the instinct to wish, trust that the seeds will find where they need to go. Grounded, quiet, real. Not Hallmark-sentimental. Tagline: "Find it. Wish it forward. Let it bloom when it's ready."
17Reference files (where the truth lives)
When you want me to load context, point me at these. Keeping this list short on purpose — these are the ones that matter most.
| File | What it is · when to reference |
|---|---|
| ASHLEY_SAFETY_RULES.md | Your guardrails. Root folder. Load first, every session. |
| SESSION-BRIEF.md | Auto-updated Sunday 8am. Current stage, blockers, priorities. Always load. |
| CLAUDE.md | MASTER folder. Project rules, naming, EOD, Green Light, Josh workflow. Always load. |
| CLAUDE-NOTES.md | MASTER folder. Deploy commands, rescore procedure, preferences, inspo library rule. |
| biz-dev-homework.html | Open items & ideas tracker. Living document. |
| DEPLOY-STATUS.md | What's 🟢 green-lit and 🔴 on hold. Check before any deploy. |
| ux-inspiration-library.html | Active design directive. Read before any UI task. |
| INSPIRATION-LIBRARY-HANDOFF.md | MASTER folder. Read if continuing inspo library work mid-build. |
| QUESTIONS-FROM-ASHLEY.md | dandyline-app/ folder. Your comms log with Josh. Read for app-build sessions. |
| JOSH-UPDATE-APRIL-12.md | MASTER folder. Strategy brief, living mirror for Josh comms. |
| SESSION-LOG.md | Root folder. Dated log of what happened each session. EOD appends here. |
| biz-dev-command-center.html | Main hub + TOC. New files should be added here. |
| biz-dev-coaching-dashboard.html | The scoreboard. Updated on "rescore." |
18Bonus recommendations (new ideas)
Things you haven't set up yet that I think will pay off. Pick the ones that resonate — not all at once.
1. A decisions log file
You don't have one yet. Right now, decisions live scattered across SESSION-LOG, homework, CLAUDE.md notes, and chat history. Create biz-dev-decisions-log.md at the MASTER folder root. Every time you say "plant a flag," I append. Result: you never re-argue a decision you already made. When Josh asks "why are we doing it this way?" you have a one-file answer.
2. A weekly "mirror" session
Once a week (Friday or whenever), run this prompt:
Weekly mirror. Read SESSION-LOG.md for the last 7 days + biz-dev-decisions-log.md + the latest SESSION-BRIEF.md.
Tell me honestly:
- What did I actually ship vs. what I thought I'd ship
- Where did I spin without moving (thinking loops, rewrites, deferrals)
- What's the single highest-leverage move for next week
- One thing I should stop doing
No pep talk. No cheerleading. This is the coach version of myself.
2. A "context fatigue" reset
Long sessions get muddy — I start carrying stale context and your brain gets crowded. When you feel it, use this:
Context fatigue reset. Summarize everything we've decided in the last hour in 5 bullets, then I'm going to start a fresh session. Save the summary to the top of SESSION-LOG.md under today's date with a "mid-session reset" marker.
3. The "three questions" pre-flight
Before starting anything non-trivial, I ask you three questions. You haven't asked me to do this consistently, but I recommend making it default for build tasks:
Before you touch anything, ask me 3 sharpening questions about today's task. The questions should be the ones whose answers would change what you build. Don't ask easy ones — ask the ones that would save us a rebuild later.
4. A "one-commit-per-session" rule
Right now, commits happen at EOD. That's fine, but if a session gets cut short before EOD, work sits uncommitted and rollback is hard. Add to CLAUDE.md: "Never leave a session without at least one commit, even if nothing is green-lit. WIP commits are fine." I'll remind you mid-session if we've crossed 45 minutes with uncommitted changes.
5. Pre-session energy check
Small habit: before I brief you, I should ask one thing — "What's your energy and time budget right now?" — and calibrate the brief accordingly. You can bake this into your openers or let me ask it automatically. Either way, it prevents me from proposing a 2-hour build on a 20-minute day.
6. A reverse brief — I tell you what I think you want
Especially powerful for long or multi-part tasks. Combine "mirror back" + "don't build yet":
Reverse brief. Before we start, tell me in 5–7 bullets what you think I want from today's session. Include:
- The outcome I care about (the real one, not just the surface ask)
- What "done" looks like
- The first 3 steps you'd take
- One thing you'd push back on if you were me
- One risk I haven't named yet
Then I'll confirm or correct — don't build yet.
7. A "founder mode" vs. "operator mode" toggle
Sometimes you're thinking about DandyLine the company (strategy, fundraising, positioning). Other times you're in the weeds (pixel placement, copy, CSS). These need different flavors of me. Try:
Switch to founder mode. For this session:
- Zoom out to company-level
- No pixel-level anything
- Push back on strategic logic, not aesthetics
- Frame everything in terms of investor narrative, category creation, moat, and growth
- Reference biz-investors.html, biz-dev-coaching-dashboard.html, and DEEP-AUDIT-FINAL-2026-04-24.md (strategy notes were migrated into homework cards on Apr 25)
When I say "back to operator mode," revert to normal build/design flow.
8. "Show me the door" — when a task doesn't fit today
Sometimes you'll ask for something and I should tell you it's the wrong moment — not because it's bad, but because it needs a decision or input you don't have right now. Tell me to flag those:
If you hit a point where this task can't progress without a decision I haven't made, a file I haven't shared, or information outside this conversation — say "show me the door" and tell me exactly what's missing. Don't fake progress. Don't guess. Don't write placeholder content hoping it's right.
9. Add this page to the Founder Tools FAB
I'll do this when you say the word — the FAB is on every key page with three links. We can bump it to four and link this playbook. Or replace one of the less-used ones. Your call.
My suggestion: keep the three existing (Dashboard / Command Center / Homework) and add this playbook as a fourth. The FAB is where your brain goes when it needs a tool. This belongs there.
19One-page cheat sheet
If you only memorize three things, make them these.
The one-line truth:
You don't have to memorize anything. Open a new session, ramble what you want, and I'll pick the best prompt from this Playbook and translate your words into it. Every time. Default behavior, no trigger phrase needed.
You don't have to memorize anything. Open a new session, ramble what you want, and I'll pick the best prompt from this Playbook and translate your words into it. Every time. Default behavior, no trigger phrase needed.
If you want to be explicit:
"Read the Playbook and pick the best prompt for what I just said." Also works: "what prompt should I use?" · "you pick"
Every session closes with:
Either "put this on hold" (creates a HOLD card) or "EOD" (runs the full wrap + deploy check). Never just walk away.
Either "put this on hold" (creates a HOLD card) or "EOD" (runs the full wrap + deploy check). Never just walk away.
When you feel stuck or drifty:
Try one of these: "mirror back" · "challenge me" · "draft three" · "what am I missing?" · "zoom out"
Try one of these: "mirror back" · "challenge me" · "draft three" · "what am I missing?" · "zoom out"