How to customize Continue.dev prompts

What this does

Modifies the system prompts and context instructions that Continue.dev sends to the language model, tailoring responses for code analysis, refactoring, or documentation tasks. After completion, prompts reflect project-specific rules and style.

Steps

1. Open the config file. Click the gear icon in the Continue sidebar or open ~/.continue/config.json directly.

2. Locate or create the systemMessage block. Add a top-level key called systemMessage with a string value describing the desired behavior.

   {
     "systemMessage": "You are a security-focused coding assistant. Always flag potential vulnerabilities and suggest mitigations. Prefer concise, annotated code examples."
   }
   

3. Define context providers for project awareness.

   "contextProviders": [
     { "name": "code", "description": "Current file context" },
     { "name": "docs", "description": "Project documentation" }
   ]
   

4. Customize slash commands (optional). Define custom slash commands under slashCommands to trigger specialized prompts.

   "slashCommands": [
     {
       "name": "review",
       "description": "Run a security review on the current file",
       "prompt": "{{{ input }}}\n\nReview the above code for security issues."
     }
   ]
   

5. Save and verify behavior. Reload VS Code and type a message or use a slash command.

Verification

Select-String -Path ~/.continue/config.json -Pattern "systemMessage|slashCommands"
# Expected: matching lines confirming custom entries exist

Common failures

• Model ignores custom prompts — The model may not respect systemMessage. Check provider documentation.
• Config not reloading — Run "Developer: Reload Window" after every save.
• JSON syntax error on save — Validate with a JSON linter before saving.
• Slash command not appearing — The command name must be lowercase.
• Context overflow — Keep systemMessage under 500 tokens for most models.

Operator checkpoint

Before treating this as solved, write down the local runtime, model or package version, hardware/backend if relevant, and the verification output. This keeps the guide useful as a Will-It-Run style decision instead of a one-off command transcript.

Related guides

• How to configure Continue.dev with Ollama
• How to use Continue.dev with custom model endpoints
• Course Local AI Fundamentals