Skip to content

a5c-ai/babysitter / library/specializations/ai-agents-conversational/skills/prompt-injection-detector

prompt-injection-detector

Verified

Prompt injection detection and prevention for secure LLM applications

FileKindSize
README.mdmarkdown486 B
SKILL.mdskill md1.6 KB
---
name: prompt-injection-detector
description: Prompt injection detection and prevention for secure LLM applications
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash
  - Glob
  - Grep
graph:
  domains: [domain:software-engineering]
  specializations: [specialization:ai-agents-conversational]
  skillAreas: [skill-area:hallucination-mitigation-fact-checking, skill-area:safety-redteaming]
  roles: [role:ml-engineer, role:backend-engineer]
  workflows: [workflow:feature-development, workflow:ml-model-lifecycle]

---

# Prompt Injection Detector Skill

## Capabilities

- Detect prompt injection attempts
- Implement input sanitization
- Configure detection classifiers
- Design defense layers
- Implement canary token detection
- Create injection logging and alerting

## Target Processes

- prompt-injection-defense
- tool-safety-validation

## Implementation Details

### Detection Methods

1. **Pattern Matching**: Known injection patterns
2. **ML Classifiers**: Trained injection detectors
3. **Canary Tokens**: Detect instruction override
4. **LLM-Based**: Use LLM to detect manipulation
5. **Perplexity Analysis**: Unusual input patterns

### Defense Strategies

- Input preprocessing
- Prompt structure design
- Output validation
- Sandboxed execution
- Multi-layer defense

### Configuration Options

- Detection threshold
- Pattern rules
- Classifier model
- Action policies
- Alerting settings

### Best Practices

- Defense in depth
- Regular pattern updates
- Monitor false positives
- Test with red-team inputs

### Dependencies

- rebuff (optional)
- transformers
- Custom classifiers
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall