Prompting

Evolve ranked novel solution ideas through multi-cycle recombination and selection

Run 2 to 8 passes through different lenses to surface hidden requirements

Generate and optimize prompts with reusable templates, standards and tools

Map complex systems with causal loops, archetypes and Meadows leverage points

Write launch prompts with exact success tests for long-running autonomous agents

Improve prompts with chain-of-thought, few-shot examples and evaluations

Test consequential decisions with a staged reasoning loop that challenges assumptions

Write optimized prompts for a specific AI tool such as Midjourney, Cursor or an LLM

Make a task deeper rather than bigger, then rewrite the prompt around it

Program LLM pipelines declaratively with DSPy and optimize prompts automatically

Constrain LLM output with regex and grammars to guarantee valid JSON or code

Write, refactor and evaluate LLM prompts with output schemas and test suites

Author and iterate on prompts in the Phoenix playground and run experiments

Open conversations with a concise, polite greeting and a formal self introduction

Rewrite a fuzzy question into a precise problem statement an agent can reason about

Mine plan review denials for patterns and draft prompt improvements as a report

Analyze problems with a database engineer's mental models and writing style

Draft copy-paste-ready /goal commands for long-running tasks in Claude Code and Codex

Stop generic AI slop in UI, copy and code with a core set of agent rules

Run agent tasks end to end with evidence-checked claims and no babysitting

Interview and build a shareable character card, from personality to avatar

Reason through complex, uncertain decisions and trade-offs in clear steps

Clean up voice dictation into a structured, concise prompt for an agent

Bundle prompts and files into oracle CLI runs with engines and saved sessions

Write instructions for agents in skills, AGENTS.md and CLAUDE.md files

Force evidence-first, calibrated answers that separate facts from assumptions

Write reusable prompt templates with structure rules and safety checks

Greet new users and share onboarding tips in a friendly first message

Question a plan one branch at a time until the approach and open decisions are clear

Keep coding autonomously, state assumptions and stop only when truly blocked

Brainstorm in parallel under different cognitive frames, then score and prune ideas

Refine prompts with proven optimization techniques for better results

Rewrite a prompt into a deeper version of the same task

List choices made during the work that you are not sure about

Make minimal prompt edits that cut false refusals from Fable models

Design a Grok bot role and write its setup prompt

Write a prompt teammates can give their agents to apply a fix or setup

Cut overthinking and force a short, practical decision

Split a tangled problem into opposing forces to find its main contradiction

Review finished work honestly and turn criticism into concrete fixes without defending

Gather firsthand facts and separate facts, inferences and unknowns before judging

Gather input from users, logs and code, then synthesize and validate a plan

Tune an agent system prompt by scoring repeated runs against a fixed task

Turn post-task feedback into an approval-gated plan for improving agent workflows

Check what a shipped AI product's system prompt says against a dated public archive

Turn your chat history, diaries and photos into a runnable digital persona

Stop and ask before acting on ambiguous requirements instead of assuming

Flag uncertainty about API endpoints, versions and signatures instead of guessing

Limit changes to exactly what was asked and skip unrequested refactors and features