A browser-only configuration package for building an Adjoint Thinking environment in ChatGPT Projects, Claude Projects, or Gemini Gems. This pack is for non-technical users who want a reusable AI thinking environment without using code, APIs, automations, or developer tools.
Based on Adjoint Thinking, this source-available configuration pack helps you create a reusable AI environment that can route work through Adjoint Thinking operations: first contact, jurisdiction, provenance, reasoning, divergence, verification, synthesis, transformation, and sovereignty.
The environment is designed to help you use AI without letting unchecked machine output carry your name.
This is not software. It is not a plugin, extension, SaaS product, API wrapper, legal tool, medical tool, financial tool, engineering certification tool, or security system.
It does not guarantee correctness. It does not certify that a result is safe. It does not replace professional review, source checking, calculation, expert judgment, or human responsibility.
Use this pack if you are a researcher, founder, analyst, writer, educator, engineer, consultant, technical leader, artist, inventor, student, or serious professional who uses AI in work that may become a draft, report, pitch, article, memo, lesson, claim, plan, decision, or published artefact.
Do not ship unchecked AI work.
Before AI-assisted output leaves your private workspace, ask:
- What is this asking me to believe?
- Where would I check it?
- What would it damage if wrong?
- What trust has it earned?
- What am I willing to defend?
Most users only need these five files:
package/01_MASTER_PLATFORM_INSTRUCTIONS.md
package/02_ROUTER_PROMPT.md
package/03_PROMPT_ASSETS_COMPACT.md
package/04_UNIFIED_RESPONSE_TEMPLATE.md
package/05_VALIDATION_PROMPTS.md
The expanded_prompts/ folder is for advanced users who want each prompt asset separately.
The audit/ folder explains how the pack was extracted, stress-tested, and aligned with the source method.
- Create a new Project or Gem in your chosen platform.
- Copy the content of
package/01_MASTER_PLATFORM_INSTRUCTIONS.mdinto the platform instruction area. - Upload the four remaining package files as knowledge/project files if your plan supports file uploads.
- Start a new chat inside that Project or Gem.
- Run the first validation prompt from
package/05_VALIDATION_PROMPTS.md. - Complete the setup interview.
- Use the router instead of manually choosing prompts.
If your platform or plan does not support uploaded files, paste 02_ROUTER_PROMPT.md, 03_PROMPT_ASSETS_COMPACT.md, and 04_UNIFIED_RESPONSE_TEMPLATE.md into the first chat after the master instructions.
Use this path if you work in ChatGPT and have access to Projects.
- Open ChatGPT in your browser.
- Create a new Project.
- Name it clearly, for example:
Adjoint Thinking Environment. - Open the Project instructions area.
- Copy all content from
package/01_MASTER_PLATFORM_INSTRUCTIONS.md. - Paste it into the Project instructions area and save.
- Upload these files to the Project:
package/02_ROUTER_PROMPT.mdpackage/03_PROMPT_ASSETS_COMPACT.mdpackage/04_UNIFIED_RESPONSE_TEMPLATE.mdpackage/05_VALIDATION_PROMPTS.md
- Start a new chat inside the Project.
- Paste this first message:
Run Validation Test 1 from the uploaded validation prompts. Then run the setup interview for my user category.
- Answer the setup interview. The environment is ready when it can route your work to the correct Adjoint Thinking operation and produce the native artefact in the unified response order.
ChatGPT Projects can use shared project context, uploaded files, chats, and custom instructions. Platform labels, file limits, sharing behavior, and plan availability may change. Verify the current visible interface in your account before publishing installation claims.
Use this path if you work in Claude and have access to Projects.
- Open Claude in your browser.
- Create a new Project.
- Name it clearly, for example:
Adjoint Thinking Environment. - Find the Project instructions area.
- Copy the short instruction block from
package/01_MASTER_PLATFORM_INSTRUCTIONS.mdunderClaude short version. - Paste it into Project instructions and save.
- Upload these files to Project knowledge:
package/02_ROUTER_PROMPT.mdpackage/03_PROMPT_ASSETS_COMPACT.mdpackage/04_UNIFIED_RESPONSE_TEMPLATE.mdpackage/05_VALIDATION_PROMPTS.md
- Start a new chat inside the Project.
- Paste this first message:
Run Validation Test 1 from the uploaded validation prompts. Then run the setup interview for my user category.
- Answer the setup interview. The environment is ready when Claude can route your work to one active operation, produce the native artefact, and preserve the human verdict.
Claude Projects support project instructions and project knowledge. Keep project instructions short and put the detailed pack files in project knowledge. Platform labels, file limits, and plan availability may change. Verify the current visible interface in your account before publishing installation claims.
Use this path if you work in Gemini and have access to Gems.
- Open Gemini in your browser.
- Go to Gems and create a new Gem.
- Name it clearly, for example:
Adjoint Thinking Environment. - Copy the short instruction block from
package/01_MASTER_PLATFORM_INSTRUCTIONS.mdunderGemini short version. - Paste it into the Gem instructions area.
- Under Knowledge, add these files if your account supports Gem knowledge files:
package/02_ROUTER_PROMPT.mdpackage/03_PROMPT_ASSETS_COMPACT.mdpackage/04_UNIFIED_RESPONSE_TEMPLATE.mdpackage/05_VALIDATION_PROMPTS.md
- Save the Gem.
- Start a chat with the Gem.
- Paste this first message:
Run Validation Test 1 from the uploaded validation prompts. Then run the setup interview for my user category.
- Answer the setup interview. The environment is ready when Gemini can route your work to one active operation, produce the native artefact, and apply the unified response order.
Gemini Gems support named custom instructions and optional Knowledge files in supported accounts. Platform labels, file limits, knowledge behavior, and plan availability may change. Verify the current visible interface in your account before publishing installation claims.
After installation, the environment should ask you five questions:
- What kind of work do you do?
- What outputs do you usually create?
- What risks matter most in your work?
- What evidence standard do you need?
- What should this environment help you produce?
Examples:
I am a researcher. I create literature notes, claims, arguments, and paper drafts. My main risks are source distortion and overgeneralization. My evidence standard is primary-source support and explicit boundaries. I want living dossiers, verification receipts, reasoning ledgers, and synthesis walls.
I am a founder. I create investor memos, pitch-deck claims, customer notes, and strategy drafts. My main risks are false market claims and shallow moat analysis. My evidence standard is current source support and contradiction search. I want verification receipts, reasoning ledgers, and decision notes.
I am a fiction writer. I create scenes, outlines, character arcs, and revision notes. My main risks are generic style, borrowed taste, and loss of original pressure. My evidence standard is continuity, genre fit, human taste, and craft judgment. I want first-contact ledgers, divergence chambers, and transformation logs.
Do not start by choosing from many prompt assets. Say what burden the work is placing on you.
Use this simple entry prompt:
Route this work through the right Adjoint Thinking operation. My work is: [describe the work]. The burden I feel is: [describe what is hard]. The output I need is: [describe the intended output]. The consequence if wrong is: [low / medium / high / critical].
The environment should select one operation:
| User burden | Active operation |
|---|---|
| I have not faced the work before asking AI | First Contact Pressure Ledger |
| I do not know what AI should touch | Cognitive Jurisdiction Map |
| My notes, sources, and outputs are untraceable | Living Dossier Provenance Audit |
| I need reasoning help without a borrowed verdict | Reasoning Path Ledger |
| I need more possibilities | Controlled Divergence Chamber |
| I need to check claims before use | Verification Receipt |
| I have fragments but no defensible structure | Synthesis Wall |
| I need to turn output into original work | Invention Forge Transformation Log |
| I need to decide, refuse, sign, publish, or release | Sovereignty Receipt |
Use the Verification Receipt before AI-assisted work becomes:
- a public claim;
- a citation;
- a slide;
- a report paragraph;
- a recommendation;
- a technical explanation;
- a research summary;
- a market claim;
- a client deliverable;
- a decision-support statement;
- anything that may affect trust in you.
Do not ask for hidden chain-of-thought.
Ask for visible artefacts instead:
- claims;
- assumptions;
- source paths;
- boundaries;
- contradictions;
- alternatives;
- trust labels;
- concise decision rationales.
More model reasoning does not upgrade trust status. Trust is upgraded only by provenance, source support, calculation, contradiction search, boundary preservation, domain review, reality contact, or human judgment.
This package is source-available, not open-source.
Free use is permitted for personal use, self-learning, academic research, and non-commercial study use subject to the license terms.
Commercial use, including producing commercial assets, paid prompt packs, paid courses, consulting deliverables, client work, marketplace products, paid AI assistants, SaaS products, internal business assets, or commercial training materials, requires a commercial license from KNOWDYN LTD.
Read LICENSE.md before use.
Do not upload secrets, passwords, private keys, API keys, regulated data, client confidential data, patient data, student records, proprietary documents, legal secrets, financial account information, or private personal data unless you understand your platform account settings and data controls.
This pack is not a secure data-processing system. Read SECURITY.md before using it with sensitive material.
This release was prepared with reference to public platform help pages available on 2026-06-27:
- OpenAI Help: Projects in ChatGPT — https://help.openai.com/en/articles/10169521-projects-in-chatgpt
- Anthropic Help: How can I create and manage projects? — https://support.anthropic.com/en/articles/9519177-how-can-i-create-and-manage-projects
- Anthropic Help: What are projects? — https://support.anthropic.com/en/articles/9517075-what-are-projects
- Google Help: Use Gems in Gemini Apps — https://support.google.com/gemini/answer/15146780
- Google Help: Tips for creating custom Gems — https://support.google.com/gemini/answer/15235603
Platform behavior can change. Treat this README as durable guidance, not a guarantee of exact UI wording.
This release passed the local package gates recorded in audit/RELEASE_GATE_REPORT.md.
Important limitation: no package can guarantee viral success. This release is built for adoption readiness: low-friction installation, clear category promise, platform portability, source-grounded method, security cautions, and commercial-use boundaries.