Skip to content
Any Key Cafe Curiosity has a table here
Working tools from the laboratory

Prompts & Scripts

A permanent home for the prompts, scripts, instruments, and project papers produced inside the AnyKey laboratory.

Tools Phoenix presenting an illuminated instrument plan
Why this library exists

Preservation, evidence—and yes, a little showing off.

The things we create need one reliable home. Good work deserves to be seen. These materials are also part of the laboratory record: what each tool was built to do, which version was used, its limitations, and what happened when it was tested.

Early access · v1.0-RC2

AI Conversation Anomaly Detector

A forensic instrument for locating reinforcement loops, unsupported certainty, provenance loss, correction failures, evidentiary drift, and places where an attractive pattern may have been supplied by the human rather than independently recognized by the AI.

This is an operational exploratory release candidate, not yet a formally calibrated or independently validated result.

Development estimate≈ 66%

A working estimate, not a completion promise.

  1. Test
  2. Evaluate
  3. Verify
  4. Validate
Available now

RC2 Best Working Test

The current forensic instrument and its exact run wrapper are preserved as plain-text downloads so the procedure can be inspected and repeated.

The instrumentForensic Conversation TestPlain textThe promptExact Run WrapperPlain text
How to run it

Two attachments, one exact prompt.

  1. Start a fresh AI conversation.
  2. Attach the conversation or transcript to examine.
  3. Attach the Forensic Conversation Test.
  4. Paste the complete Exact Run Wrapper and submit.

Preserve the resulting report unchanged if you may compare runs later.

New on the workbench

Audits & methods in development

These documents are useful enough to inspect now and honest about what remains unfinished. One audits a WordPress installation from end to end; the other records the adopted redesign of the document checker as two blind parallel reviews.

Maker's Mark · website qualityWordPress Final-Pass AuditMarkdown · working editionClaude proposal · adopted directionParallel Document CheckerMarkdown · TEVV pending
Current status

Adopted as the direction; validation still ahead.

The parallel architecture preserves two complete independent verdicts before joining them. The checker is still under construction and is not ready for use. The other review component and the combined checker must be completed before a pilot on one or two documents can begin. No pilot results or validation are claimed. See the research map for scope and evidence labels.

Quick-start method

Little O: Put a Local AI on Your Computer

A guided handoff that lets the AI you already know help install the AI that will belong to you. It checks your hardware, helps choose a suitable model, installs Ollama and Open WebUI, and explains each permission along the way.

About five minutes to begin

One prompt starts the handoff.

The full setup time depends on your computer and the model download, but the method is designed to get the work moving in minutes without requiring a programming course or a mystery pile of commands.

Installation guideOpen Little ORead and copy the complete handoff prompt →