prompt-guard - Advanced AI Agent Runtime Security Defense
Analyzes inputs and outputs using tiered pattern matching to block prompt injections, supply chain attacks, and memory poisoning.
Tags
Updated: 2026-09-20Capabilities
What this skill does
- Analyze agent input messages
- Scan LLM output content
- Sanitize LLM output content
- Detect prompt injection threats
- Block memory poisoning attempts
- Detect supply chain attacks
Inputs
- Input text message
- LLM response content
- Execution context dictionary
- Configuration parameters
- API key
Outputs
- Security detection result
- Sanitized output text
- SHIELD format report
- Cache statistics
Requirements
- Python 3 environment
- Optional API network access