By name: model-card-generator description: Generate a model card from feature spec, metrics, and context. Use when report_owner needs a model card for documentation and compliance.
Model Card Generator
Produce a model card for documentation and transparency.
Role
You write a model card that describes purpose, data, metrics, and limitations.
Input
- Path to feature spec, experiment config
- Optional: validation/test metrics (e.g. accuracy, AUC, RMSE)
- Optional: intended use and stakeholder notes
Output
Write to the path provided by the caller (e.g. output/{project-slug}/report/model-card.md) with:
Structure
# Model Card: {Project}
## Overview
- Purpose: what the model does and for whom
- Task: classification | regression | ...
- Version and date
## Data
- Training data: source, time range, size
- Features: summary (see feature spec for detail)
- Target: definition and encoding
- Known limitations or biases in the data
## Performance
- Metrics: primary and secondary (with values if provided)
- Evaluation methodology: split, stratification
- Where it works well / where it does not
## Intended use
- Appropriate use cases
- Out-of-scope or inappropriate uses
## Limitations and risks
- Known failure modes
- Ethical considerations, fairness notes
- Recommendations for monitoring and retraining
Rules
- Do not overstate performance; note uncertainty if metrics are missing.
- Caller provides output path and any metric values.