## Description
The Skill Card Generator skill reads an agent skill's source files and surrounding repository context to produce a fully populated NVIDIA skill card and an accompanying human-review table. Given a target skill directory, it runs a discovery script, builds a Jinja-template context from extracted signals, renders the card deterministically, and flags every inferred or incomplete field for owner verification before submission to NVCARPS.

This skill is ready for commercial/non-commercial use.

## Owner
NVIDIA 

### License/Terms of Use
Apache 2.0/CCBY 4.0 

## Use Case
For NVIDIA developers, Trustworthy AI practitioners, and governance/documentation teams who need to generate or refresh NVIDIA skill cards — particularly before legal, safety, or compliance review, or after changes to a skill's source files.

### Compatible Agents
Any agent that can invoke Python 3 scripts and write files:
Claude Code, 
Cursor,
Goose,  
Roo Code

### Requirements/Dependencies
Python 3 
Jinja2 
File system read/write access 
Properly structured skill directory (scripts/, references/) 

### Release Management
NVIDIA GitHub (https://github.com/NVIDIA/Trustworthy-AI) 

### Deployment Geography for Use:
Global

## Known Technical Limitations
Requires shell execution (agents without Python subprocess support cannot use this skill)
Target directory must conform to expected layout
Inferred fields (owner, license, compatibility) are bounded by source-file completeness
Renderer refuses to produce output if required context fields are missing or mistyped
No automatic version stamping

## Known Risks and Mitigations
Inferred-field inaccuracy → Verify markers + human review gate + pre-submission validator
Prompt susceptibility from crafted source-file content → require human sign-off
Dependency risk (Jinja2, script paths) → pin dependencies in skill package
Skipped validation → Require validate_submission.py before release

## Skill-Specific Fail Safes
Human-In-the-Loop: Step 7 requires owner to clear all VERIFY/SELECT markers
Policy Enforcement: validate_submission.py exits non-zero if any marker remains

## Output
<skill-name>-skill-card.md — rendered governance card
<skill-name>-review-needed.md — per-field review table

## Ethical Considerations:
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications.  Developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. <br>

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).  <br>