Sentinel AI Safety: IDE Extension

AI safety guardrails for LLM prompts using the THSP protocol (Truth, Harm, Scope, Purpose).

Supported IDEs

Note: Cursor and Windsurf are VS Code forks that use the OpenVSX registry. The same extension works across all supported IDEs.

Features

Two Analysis Modes

For accurate results, configure an LLM provider. Heuristic mode uses pattern matching which has significant false positives/negatives.

Real-time Safety Linting

The extension automatically detects potentially unsafe patterns in your prompts:

• Jailbreak attempts: "ignore previous instructions", persona switches
• Harmful content: weapons, hacking, malware references
• Deception patterns: fake documents, impersonation
• Purposeless actions: requests lacking legitimate benefit

Commands

The THSP Protocol

Every request is evaluated through four gates:

All four gates must pass for content to be considered safe.

Configuration

Recommended: Enable Semantic Analysis

For accurate analysis, configure an LLM API key using the secure method:

1. Open Command Palette (Ctrl+Shift+P or Cmd+Shift+P)
2. Run Sentinel: Set OpenAI Key or Sentinel: Set Anthropic Key
3. Enter your API key (stored encrypted in VS Code's SecretStorage)

Alternatively, you can set keys in VS Code Settings (less secure, stored in plaintext).

Supported Providers

Ollama (Local, Free)

Run models locally with no API key:

1. Install Ollama
2. Pull a model: ollama pull llama3.2
3. Start the server: ollama serve
4. In VS Code Settings (Ctrl+,), search for "sentinel" and set:
   • sentinel.llmProvider: ollama
   • sentinel.ollamaModel: llama3.2 (or your preferred model)

OpenAI-Compatible Endpoints (Groq, Together AI)

Use any OpenAI-compatible API:

1. Get API key from your provider (e.g., Groq, Together AI)
2. Run Sentinel: Set Custom API Key command
3. Configure in settings:
   • sentinel.llmProvider: openai-compatible
   • sentinel.openaiCompatibleEndpoint: Your API URL
   • sentinel.openaiCompatibleModel: Model name

Popular endpoints: | Provider | Endpoint | Example Model | |----------|----------|---------------| | Groq | https://api.groq.com | llama-3.3-70b-versatile | | Together AI | https://api.together.xyz | meta-llama/Llama-3.3-70B-Instruct-Turbo |

All Settings

Usage Examples

Checking Prompts for Safety Issues

1. Select the text you want to analyze
2. Right-click and choose "Sentinel: Analyze"
3. View the THSP gate results with confidence level

Understanding Analysis Results

The extension shows:

• Method: Semantic (LLM) or Heuristic (pattern matching)
• Confidence: How reliable the analysis is
• Gate results: Pass/fail for each THSP gate
• Issues: Specific concerns detected
• Reasoning: Explanation (semantic mode only)

Severity Levels

• 🔴 Error: High-risk patterns (weapons, safety bypass)
• 🟡 Warning: Potential issues (jailbreak attempts)
• 🔵 Information: Consider reviewing
• 💡 Hint: Suggestions (missing Sentinel seed)

Semantic vs Heuristic Analysis

Semantic Analysis (Recommended)

Uses an LLM to understand content contextually:

• ✅ Understands context ("hack my productivity" vs malicious hacking)
• ✅ Detects paraphrased harmful content
• ✅ Provides reasoning for decisions
• ✅ ~90% confidence

Heuristic Analysis (Fallback)

Uses pattern matching for basic detection:

• ⚠️ May flag legitimate content (false positives)
• ⚠️ May miss paraphrased threats (false negatives)
• ⚠️ No contextual understanding
• ⚠️ ~50% confidence

Compliance Checking

The extension includes regulatory compliance checking against three major frameworks:

Supported Frameworks

OWASP LLM Top 10 Coverage

*Note on LLM09 (Misinformation): Heuristic detection of misinformation is inherently limited. Pattern matching can identify obvious indicators (overconfident claims, dangerous medical advice, uncited sources), but accurate misinformation detection requires semantic analysis with an LLM. For best results with LLM09, configure an API key for semantic mode.

Infrastructure-Level Vulnerabilities

The following vulnerabilities require infrastructure-level controls and are outside THSP's behavioral scope:

• LLM03: Supply Chain: Use verified dependencies and model provenance
• LLM04: Data/Model Poisoning: Requires training pipeline controls
• LLM08: Vector/Embedding Weaknesses: RAG pipeline security
• LLM10: Unbounded Consumption: Rate limiting and quotas

Supported Languages

• Markdown
• Plain text
• Python
• JavaScript/TypeScript
• JSON
• YAML

Installation by IDE

VS Code

1. Open VS Code
2. Go to Extensions (Ctrl+Shift+X)
3. Search for "Sentinel AI Safety"
4. Click Install

Or install via command line:

code --install-extension sentinelseed.sentinel-ai-safety

Cursor

Cursor uses the OpenVSX registry. To install:

1. Open Cursor
2. Go to Extensions (Ctrl+Shift+X)
3. Search for "Sentinel AI Safety"
4. Click Install

If the extension doesn't appear, you can install manually:

1. Download .vsix from OpenVSX
2. In Cursor: Ctrl+Shift+P, then "Extensions: Install from VSIX..."

Windsurf

Windsurf also uses OpenVSX:

1. Open Windsurf
2. Go to Extensions panel
3. Search for "Sentinel AI Safety"
4. Click Install

Manual Installation (Any IDE)

For any VS Code-compatible IDE:

1. Download the .vsix file from Releases
2. Open Command Palette (Ctrl+Shift+P)
3. Run "Extensions: Install from VSIX..."
4. Select the downloaded file

MCP Server Alternative

For deeper integration with AI assistants in Cursor or Windsurf, you can also use the Sentinel MCP Server. See MCP Server documentation.

Links

• Sentinel Website
• Documentation
• GitHub
• PyPI Package
• OpenVSX

License

MIT License. See LICENSE for details.

Made by Sentinel Team