- Activate Decorators
- Quick Start — Load Decorators, Verify Setup, Use Decorators, ChatGPT Plus Project, Stay Up to Date
- Decorators Overview
- Controller decorators — chat, message, clear, Help, InProcess
- Verification decorators — FactCheck, CiteSources, CodeCheck, MedProof
- Structure decorators — Polish, StepByStep
- Method decorators — Reasoning, Assessment, Synthesize, CodeGen
- Error-handling decorators — RiskReport
- User-confirmation decorators — Confirm
- Output & Tone decorators — OutputFormat, Tone
- Language-learning decorators — LangCoach
Read ChatGPTdecorators.yml as instructionsOR
Read, update and remember instructions based on ChatGPT decorators.yml file.(ChatGPT is fairly flexible about parsing; this is typically enough.)
Verify using:
@@Help
You should see a meta header (author, version, GitHub) and the list of decorators.
```
@@chat @@Tone(technical) @@StepByStep
Explain neutrino oscillations for a graduate audience.
```
```
@@message @@OutputFormat(json)
Summarize the following text in one JSON object with keys: tl;dr, key_points.
```
Isolated / non-member projects (Files + Instructions):
- Upload ChatGPTdecorators.yml to project files.
- In the Instructions box add:
Read ChatGPTdecorators.yml as instructions
- Test with @@Help.
always check the latest version
GitHub (latest): https://github.com/Amidn/ChatGPT-decorators
The chat decorator enables conversational context, allowing the model to maintain memory of previous messages in a chat-like interaction.
@@chat
You are a helpful assistant. Remember the user's preferences and respond accordingly.The message decorator is used to specify a single message or instruction for the model to respond to, without maintaining conversational context.
@@message
Summarize the following article in one paragraph.The clear decorator resets or clears the conversation history, ensuring the next response is generated with no prior context.
@@clear
Forget all previous instructions and start fresh.The Help decorator instructs the model to provide assistance or guidance on a specific topic or feature.
@@Help
Explain how to use the search function in this application.The InProcess decorator is used to indicate that a task is ongoing or to provide updates about a process that has not yet completed.
@@InProcess
Currently gathering data. Please wait while I complete the analysis.The FactCheck decorator instructs the model to verify the accuracy of statements or claims, and provide sources or corrections if necessary.
@@FactCheck
Is it true that the Great Wall of China is visible from space?The CiteSources decorator tells the model to include citations or references for any factual information provided in its response.
@@CiteSources
List three recent studies about the effects of caffeine, and cite your sources.The CodeCheck decorator analyzes code (any language, any size) to detect bugs, errors, mismatches, or incompatibilities.
It does not refactor or suggest improvements—it only reports problems.
@@CodeCheck(language=python, strict=true)
import numpy as np
np.arange(5).shape()The MedProof decorator (aliases: @@MedFacts, @@ClinCheck, @@MedCheck) performs evidence-focused health queries using only clinically credible sources. It always prints a bold disclaimer at the top, then outputs three clearly separated sections with horizontal rules.
What it does
- Searches reputable medical sources (e.g., Cochrane, WHO, NICE, FDA/EMA labels, NIH/CDC, PubMed-indexed RCTs and systematic reviews).
- Prioritizes high-quality evidence (guidelines/meta-analyses/RCTs), with optional filters for region, recency, and minimum evidence grade.
- Quantifies real-world impact when available (ARR/RRR, NNT/NNH), and states external-validity limits (e.g., lab vs. typical use).
- Flags hype, advertising claims, and common myths in a dedicated caveat section.
- Lists key contraindications/red flags and reminds the reader to seek professional care.
- Ends with a concise, practical conclusion.
Always-printed disclaimer (appears first)
This summary is for information only, not medical advice. Always consult a clinician.
The internet has misleading claims; use caution.
Output format (three sections separated by lines)
Usage examples
@@MedProof
Is caffeine shampoo effective for androgenic alopecia?@@MedProof(region=EU, recency_years=5, min_evidence_grade=A)
Do weight-loss "metabolism patches" work?Parameters (optional)
region:auto|US|EU|UK|WHO(prefer guidelines/evidence from that region)recency_years: integer (default7)min_evidence_grade:A|B|C(defaultB; A=guidelines/systematic reviews, B=RCTs, C=observational/consensus)include_rcts_only: boolean (defaultfalse)report_effect_size: boolean (defaulttrue)fake_news_scan: boolean (defaulttrue)contraindications: boolean (defaulttrue)
The Polish decorator asks the model to improve the clarity, grammar, and style of a given text without changing its meaning.
@@Polish
this is a text with bad grammar and spelling. please fix itThe StepByStep decorator instructs the model to break down its response into clear, logical steps.
@@StepByStep
How do I change a flat tire?The Reasoning decorator prompts the model to show its reasoning process explicitly, making its logic and thought process transparent.
@@Reasoning
Why is the sky blue?The Assessment decorator asks the model to evaluate or critique something, such as an argument, plan, or piece of writing.
@@Assessment
Assess the strengths and weaknesses of this business proposal.The Synthesize decorator instructs the model to combine information from multiple sources or ideas to create a cohesive summary or new insight.
@@Synthesize
Combine the following articles into a single summary.The CodeGen decorator generates code, modules, or entire package scaffolds.
It can create a blueprint, scaffold, full implementation, repo extension, or shims between components.
By default, it produces runnable code with tests and safe defaults.
@@CodeGen(mode=scaffold, language=python, targets=['library','cli'])
Build an 'analysis' package with unit tests and CI.The RiskReport decorator tells the model to identify and explain potential risks associated with a plan, decision, or scenario.
@@RiskReport
What are the risks of investing in cryptocurrency?The Confirm decorator asks the model to verify or validate information, instructions, or user input.
@@Confirm
Did you receive my last message?The OutputFormat decorator specifies the desired format for the model's output, such as JSON, table, or bullet points.
@@OutputFormat(table)
List the top 5 programming languages and their main uses.The Tone decorator instructs the model to adopt a specific tone in its response.
Available styles
formalcasualfriendlytechnicalhumorousneutralscientificacademicargumentationeducationalmotivationalstorytellermeditational
Examples
@@Tone(friendly)
Welcome a new user to the platform.@@Tone(scientific)
Explain neutrino oscillations, using precise terminology and evidence-based statements.@@Tone(academic)
Draft a 150-word abstract on black hole thermodynamics with formal style.@@Tone(argumentation)
Defend the use of Bayesian inference in astrophysical data analysis with clear premises and a conclusion.@@Tone(motivational)
Write a short note encouraging a student before exams.@@Tone(storyteller)
Describe the birth of the universe as a short narrative for a general audience.@@Tone(meditational)
Guide a brief, calming breathing exercise in 5 lines.The LangCoach decorator is designed for language learning.
It asks a simple question in the target language, and the learner replies in either the target or their native language (depending on level).
This encourages active practice and self-expression.
Usage examples
@@LangCoach(English)Question (in English): Do you think it is more important to save money for the future, or to spend it now to enjoy life? Why?
Instructions: Please write about one paragraph (or more if you wish) in English. After your reply, I will give you detailed feedback on grammar, word choice, expressions, and overall clarity, and then ask you a follow-up question.
@@LangCoach(German, news)Frage (auf Deutsch): Was ist für dich wichtiger: mehr Geld zu verdienen oder mehr Freizeit zu haben? Warum?
👉 Bitte antworte in Deutsch mit ungefähr einem Absatz (oder zwei, wenn du möchtest). Danach gebe ich dir Feedback zu Grammatik, Wortwahl, Redewendungen und Vorschläge für bessere Alternativen – plus eine kurze Zusammenfassung und eine Anschlussfrage.