Build a claim-to-source ledger
Makes an article, pitch or report auditable line by line.
A growing, practical AI prompt library
Find complete prompts for real tasks, understand what each one does, replace the variables and use it immediately.
New: prompts for messy reality
Turn claims, files, screenshots and risky offers into answers you can inspect. These prompts show what AI can and cannot verify.
Copy, paste, smile
These short prompts work in a chat you already use. For personal questions, the AI should say what it actually knows, not pretend it can see your whole life.
Find your starting point
Choose a format, then explore its topics. Every prompt has its own page.
Browse the library
Makes an article, pitch or report auditable line by line.
Defines evidence thresholds that make a future review worthwhile.
Compares the original reasons with new evidence without hindsight theatre.
Separates intention, tool action, server result and user-visible outcome.
Checks every important claim against evidence the AI can actually see.
Converts source-backed research into options, trade-offs and a recommended test.
Scopes a code edit, preserves user changes and verifies the requested behaviour.
Makes a schema-driven extraction agent flag uncertainty instead of filling gaps.
Tests a full customer path and records evidence at each step.
Researches and drafts a measured content calendar but never auto-posts.
Sorts messages and drafts replies without sending them.
Classifies a request and prepares an evidence-based response for human review.
Defines delegation boundaries and an evidence-rich handoff.
Checks a condition on a schedule without noisy unchanged updates.
Adds hard step, time and spend limits before an agent loops.
Defines what an agent should remember, forget and never store.
Turns approved documents into a maintained source map rather than a context dump.
Checks whether an answer is genuinely grounded in the sources it claims.
Prevents a helpful assistant from silently sending, buying or deleting.
Diagnoses an agent run without turning guesses into fixes.
Creates normal, edge, adversarial and failure tests with observable pass criteria.
Sorts tools into read, draft, approval and prohibited capabilities.
A fill-in-the-blanks operating contract for a narrow agent.
Need an AI worker, not one answer?
Define tools, authority, proof, persistence and safe stop conditions.