Prompts that match how you think.
Most prompt packs are landfill: a thousand untested one-liners with no method behind them. These are the opposite: eleven prompts per pack, each tied to a real workflow, each carrying its failure mode: the exact, observed way AI gets that job wrong, and the move that catches it. Buy the pack that matches your grain.
The whole market competes on volume. We compete on trust.
Few, each excellent
Eleven prompts per pack, not a thousand. Every one tied to a real workflow with a known output. Nothing in here is filler.
A failure log, not a label
Every prompt carries an observed failure: the exact moment the output looked right and wasn't, plus the counter-move that catches it. An AI can copy the prompts in ten minutes; it can't copy the error log.
Matched to how you work
You don't browse a thousand prompts hoping one fits. You buy the pack for how you think: the systems-builder's, or the truth-seeker's.
Pick the one that matches your grain.
Each is a focused operating manual for one kind of thinker: eleven prompts, every one with its guardrail, all templated for your domain. Not sure which is you? Most people know within the first line.
The Architect's Toolkit
"You think top-down, structure-first, and you reward depth. AI builds the frame, you guarantee it's sound."
- Architecture Decision Record
- System Dependency Map
- Technical / Product Spec
- Decision Briefing Document
- Runbook / SOP Generation
- QA Checklist · Project Risk Audit
- Mental Model · Team Standards Doc
- + 2 more, each with its counter-move
Technical / Product Spec
When something needs building and you want the requirements unambiguous before work starts.
Draft a spec for: [WHAT'S BEING BUILT]. Include: problem statement, goals, non-goals, requirements, constraints, open questions, and success criteria. Mark every assumption explicitly. List what you'd need to know to remove each open question.
The Sage's Toolkit
"You know why, not just what. Let AI do the reading, you do the judging."
- Evidence Synthesis + disagreement mapping
- Claims & Assumptions Audit
- Targeted Deep Dive
- Trend Signal Synthesis
- Steelman the Opposing View
- Literature Map · Source Triage
- Confidence Calibration
- + 3 more, each with its counter-move
Evidence Synthesis (confidence-rated)
When you need to know what the evidence actually says, not the first plausible answer.
Research question: [QUESTION]. Synthesize what the evidence actually shows. Distinguish well-supported claims from commonly repeated assumptions. Rate each claim's evidence quality: strong / moderate / weak. Note what's missing or unstudied. Then flag 2 to 3 things most practitioners get wrong about this.
The Catalyst's Toolkit
"You read the room, rally the doubters, and make people want to do the work. AI drafts the message — you add the warmth and the read of the room."
- Audience-Tailored Message
- The Resistance Map
- Persuasive Narrative / Pitch Spine
- Stakeholder Buy-In Map
- Difficult Conversation Prep
- Meeting → Aligned Action · Rollout Message
- Tough-Question Prep · Recognition That Lands
- + 2 more, each with its counter-move
Audience-Tailored Message
When you have to tell several groups the same news, and each one cares about something different.
I need to communicate this to several audiences: [THE NEWS/DECISION]. Audiences: [LIST EACH GROUP — e.g. exec team, affected team, customers]. For each, write a short version that leads with what THEY care about. For each, name the one thing they most want to know and the one thing they're most worried about.
Three free prompts — one from each toolkit.
Take these and use them today. Each is a real prompt from its pack, with the guardrail that catches where AI quietly gets the job wrong. Want more? Every pack shows two additional prompts free on its preview page.
Technical / Product Spec
Draft a spec for: [WHAT'S BEING BUILT]. Include: problem statement, goals, non-goals, requirements, constraints, open questions, and success criteria. Mark every assumption explicitly. List what you'd need to know to remove each open question.
Evidence Synthesis
Research question: [QUESTION]. Synthesize what the evidence actually shows. Distinguish well-supported claims from commonly repeated assumptions. Rate each claim's evidence quality: strong / moderate / weak. Note what's missing or unstudied. Then flag 2–3 things most practitioners get wrong about this.
Audience-Tailored Message
I need to communicate this to several audiences: [THE NEWS/DECISION]. Audiences: [LIST EACH GROUP — e.g. exec team, affected team, customers]. For each, write a short version that leads with what THEY care about. For each, name the one thing they most want to know and the one thing they're most worried about.
Buy a pack. Or get all three and save.
- 11 workflow-mapped prompts
- Every prompt's guardrail included
- Interactive web page (copy-paste buttons) + PDF
- Individual-use license
Checkout is being connected. Email to reserve The Architect. You’ll be first in line.
- 11 workflow-mapped prompts
- Every prompt's guardrail included
- Interactive web page (copy-paste buttons) + PDF
- Individual-use license
Checkout is being connected. Email to reserve The Sage. You’ll be first in line.
- 11 workflow-mapped prompts
- Every prompt's guardrail included
- Interactive web page (copy-paste buttons) + PDF
- Individual-use license
Checkout is being connected. Email to reserve The Catalyst. You’ll be first in line.
- All 33 prompts, all three packs
- Every guardrail included
- All three as interactive web pages + PDFs
- Individual-use license
Checkout is being connected. Email to reserve the full bundle. You’ll be first in line.
No-risk: if a pack doesn't sharpen how you work with AI, reply to your receipt within 14 days for a full refund. The guardrails are the point: if they don’t earn their keep, you don’t keep the charge. Checkout runs through Gumroad; buying grants the individual-use license under the terms of sale.
More than a list. An operating manual.
An identity intro
A one-page "how to think like this archetype", so you're buying an operating system, not a file of strings.
11 workflow prompts
Each with when to use it, the copy-paste prompt with [BRACKETS] to fill, and its counter-move.
An observed failure on each
The exact way AI gets that task wrong, named from real use, plus the move that catches it. The one thing no other pack on the market gives you.
A path to the done-for-you
When the decision's big enough that you'd rather we ran it, each pack points to the matching Ousios workflow.
Quick answers.
Which pack is me?
The Architect if your instinct is to build the structure: specs, decisions, systems, runbooks. The Sage if your instinct is to check what's true: research, synthesis, finding where the experts actually disagree. The Catalyst if your instinct is to bring people with you: buy-in, narrative, and rolling out change that sticks. Most people lean one way and know it immediately. If you do more than one, the $99 three-pack is the obvious call.
Isn't this just prompts I could write myself?
You could, and the prompt is the cheap part. What you're buying is the guardrail on each one: the specific, hard-won way the AI tends to get that exact task wrong, so you catch it before it costs you. That's the part you can't easily write yourself, because it comes from running the workflow, not imagining it.
What format do I get?
Both: an interactive web page with one-tap copy buttons on every prompt and counter-move, an at-a-glance index, a worked before/after example, and a one-page cheat-sheet, plus a clean, designed PDF for reading or printing. Instant download right after checkout, through Gumroad. Use the prompts in any AI tool: Claude, ChatGPT, Gemini, whatever you run.
Will these work in my industry?
Yes, with one honest caveat. Every prompt is templated with [BRACKETS]; the method is constant and the subject is a variable. For regulated work (clinical, legal, financial), several prompts include an "applicable standards" bracket: fill it, and the output stays audit-aware instead of confidently generic. That bracket is the difference between a prompt that looks right and one that's safe to act on.
Eleven prompts each. Every one honest about where AI fails.
That honesty is the whole point: a fast, tireless assistant whose work you can actually trust, because you know exactly where to check it.