how-to-use-ai
llms.txt0 starsChanged 7 days agoMITView on GitHub
Evidence-based guide to using AI well: how to instruct models (role, context, instruction, format, constraints, examples) and how to prevent the four documented failure modes (misunderstanding, omission, fabrication/hallucination, sycophancy). Includes installable Agent Skills, a drop-in AGENTS.md rules file, a prompt library, a runnable eval, and sourced research reports (460+ sources, checked 2026-11-09).
In this repository11 other files it gives its agents.
- drop-in/AGENTS.mdAGENTS.md
- clarify-firstSkill
- coverage-checkSkill
- fresh-eyes-reviewSkill
- fresh-start-handoffSkill
- grounded-answerSkill
- honest-critiqueSkill
- learning-tutorSkill
- prompt-improverSkill
- task-specSkill
- verify-claimsSkill
The file
Files
# how-to-use-ai > Evidence-based guide to using AI well: how to instruct models (role, context, instruction, format, constraints, examples) and how to prevent the four documented failure modes (misunderstanding, omission, fabrication/hallucination, sycophancy). Includes installable Agent Skills, a drop-in AGENTS.md rules file, a prompt library, a runnable eval, and sourced research reports (460+ sources, checked 2026-11-09). ## Install - [Skills (Agent Skills format)](https://github.com/ji-hun-git/how-to-use-ai/blob/main/skills/README.md): 10 skills; Claude Code plugin, `npx skills add ji-hun-git/how-to-use-ai`, or manual copy - [Drop-in rules file](https://github.com/ji-hun-git/how-to-use-ai/blob/main/drop-in/AGENTS.md): 18 evidence-linked rules for AGENTS.md / CLAUDE.md - [Custom instructions for chat apps](https://github.com/ji-hun-git/how-to-use-ai/blob/main/drop-in/custom-instructions.md): short and long versions ## Guides - [Mindset](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/01-mindset.md): how to think about AI, jagged capabilities, when not to use it - [The six elements of a prompt](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/02-prompt-anatomy.md): role, context, instruction, format, constraints, examples - [Instructing with examples](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/03-examples-few-shot.md): few-shot rules and pitfalls - [Failure modes](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/04-failure-modes.md): misunderstanding, omission, fabrication, sycophancy; causes, prevention, detection, 2026 measurements - [Models and reasoning](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/05-models-and-reasoning.md): choosing models and tools, reasoning models, habits to drop - [Workflows](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/06-workflows.md): context engineering, five patterns, playbooks, memory - [AI coding](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/07-ai-coding.md): agentic engineering, done criteria, verification, 2026 tool changes - [Verification and evals](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/08-verification-and-evals.md): verification ladder, evaluating prompts - [Security and privacy](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/09-security-privacy.md): prompt injection, lethal trifecta, local laws example - [Myths vs. evidence](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/10-myths-vs-evidence.md): which popular tips fail in controlled studies - [Where instructions should live](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/11-skills-and-agent-rules.md): prompt vs rules file vs skill vs hook - [Writing decision rules](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/12-writing-decision-rules.md): condition → action → exception rules, exceptions and human sign-off, testing that rules fire - [Methods and the papers behind them](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/14-methods-and-papers.md): 73 prompting, grounding, verification, agent and context methods graded for 2026, with 192 sources - [Verifying agent runs](https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/13-verifying-agent-runs.md): three levels of done, five silent failures, test sets, tracing causes, change logs, measuring benefit ## Practical - [Prompt library](https://github.com/ji-hun-git/how-to-use-ai/blob/main/prompts/README.md): 30 prompts by task - [Prompt clinic](https://github.com/ji-hun-git/how-to-use-ai/blob/main/examples/08-prompt-clinic.md): 12 common prompts diagnosed and fixed - [Templates](https://github.com/ji-hun-git/how-to-use-ai/tree/main/templates): 10 annotated templates - [Cheatsheet](https://github.com/ji-hun-git/how-to-use-ai/blob/main/CHEATSHEET.md): one-page summary ## Evidence - [Evidence report](https://github.com/ji-hun-git/how-to-use-ai/blob/main/research/evidence-report.md): sourced report with verification status per claim - [2026 update report](https://github.com/ji-hun-git/how-to-use-ai/blob/main/research/2026-update-report.md): what changed in 2026 and current-model measurements ## Optional - [How this repo compares](https://github.com/ji-hun-git/how-to-use-ai/blob/main/COMPARISON.md) - [Curated resources](https://github.com/ji-hun-git/how-to-use-ai/blob/main/resources/awesome.md) - [Runnable eval](https://github.com/ji-hun-git/how-to-use-ai/blob/main/evals/README.md) - [Skill eval results](https://github.com/ji-hun-git/how-to-use-ai/blob/main/evals/results/README.md): with-vs-without results for all 10 skills on two suites (36 cases), with raw answers
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

