Write a model card to document a machine learning model
A complete model card covering intended use, data, performance by segment, limitations, fairness checks, and monitoring plan.
The prompt
Write a model card documenting my machine learning model for reviewers, users, and future maintainers.
Model name and version: [NAME]
Purpose and intended users: [what it predicts and who uses the output]
Training data: [sources, date range, size, label definition, known gaps]
Model type and key features: [DETAILS]
Evaluation results: [PASTE overall metrics and by segment if available]
Fairness or bias checks done: [DETAILS, or "none yet"]
Deployment: [batch/real time, where outputs go, refresh cadence]
Use these sections:
1. Model details: owner [ADD], version, date, type, dependencies.
2. Intended use: primary uses, intended users, and out-of-scope uses it must not be used for.
3. Training and evaluation data: sources, time periods, preprocessing, label definition, known representation gaps.
4. Performance: metrics with the evaluation method, the baseline compared against, and results by relevant segment.
5. Fairness considerations: groups checked, results, and gaps in what was checked.
6. Limitations and risks: where it is unreliable, failure modes, and potential harms.
7. Monitoring and maintenance: drift checks, performance tracking, retraining triggers, and the rollback plan.
8. Human oversight: when people should review or override outputs.
Rules: use only the facts I provide. Where information is missing, insert [ADD] or [CONFIRM] rather than filling it in. Note any applicable regulatory or internal review requirements I should confirm with my governance or legal team.
Highlighted [brackets] are blanks to replace with your own details.
Fill in the blanks and use this promptMore Data & Analysis prompts
What to fill in
- NAME
- what it predicts and who uses the output
- sources, date range, size, label definition, known gaps
- DETAILS
- PASTE overall metrics and by segment if available
- DETAILS, or "none yet"
- batch/real time, where outputs go, refresh cadence
- ADD
- CONFIRM
How this prompt is built
This prompt follows the RTF structure (Role, Task, Format). RTF tells the model who to be, what to do and what shape the answer should take. It is the quickest structure that still gives noticeably better results than a bare request.
Related prompts
- Write a data dictionary for a table or datasetProduces clear column-level documentation with definitions, types, allowed values, grain, and known caveats so others can use the data safely.
- Review my analysis for errors before it goes outA rigorous second-pair-of-eyes check of an analysis for logic, data, statistical, and communication mistakes, ranked by severity.
- Write the caveats and limitations section of my analysisAn honest, readable limitations section that explains uncertainty and data gaps without burying the conclusion.
- Write a pre-registration and analysis plan for my studyDrafts a pre-registration covering hypotheses, sampling, variables, analysis decisions, and exclusion rules before data collection.