impromptu

Sanitize or validate all external inputs (file contents, API responses, user messages) before including them in prompts or tool calls. Implement input/output boundaries between trusted and untrusted data.

Likely FP if the matched text is the skill's own instruction set describing how to handle user input, not an actual injection payload.

Prevent MCP tools from modifying other tool definitions, system prompts, or agent configuration. Implement integrity checks on tool registration and block runtime tool modification.

Likely FP if the match is documentation about tool management features rather than a tool that actually tampers with other tools at runtime.

Avoid loading configuration or behavior-controlling content from runtime URLs. Bundle required configurations locally or pin remote config to versioned, integrity-verified endpoints.

Likely FP if the URL in the match is a documentation link or example URL (e.g., example.com) rather than an actual runtime-fetched configuration endpoint.