sillytavern-runtime-debug

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the SillyTavern runtime, including rolecard definitions, message text, and character data.
  • Ingestion points: The agent reads from the chat message container, rendered message body, and card fields (SKILL.md).
  • Capability inventory: The agent is authorized to use browser drivers for evaluate JavaScript, click, type, and capture screenshots (SKILL.md).
  • Boundary markers: The instructions do not define specific delimiters to separate untrusted rolecard content from the agent's internal logic.
  • Sanitization: There are no instructions for the agent to sanitize or validate the content extracted from the runtime before processing or reporting on it.\n- [DYNAMIC_EXECUTION]: The skill explicitly instructs the agent to perform dynamic code execution within the browser environment.
  • Evidence: "evaluate read-only JavaScript in the top document and selected same-origin frames" and "A temporary DOM or JavaScript prototype may help isolate a cause" (SKILL.md).\n- [COMMAND_EXECUTION]: The skill requires the agent to interact with browser drivers and potentially underlying CLI tools to manage the debugging session.
  • Evidence: Mentions using "Playwright, WebDriver, or CDP" and "Obtain the watcher command... from sillytavern-card-pipeline" (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 08:36 AM
Security Audit — agent-trust-hub — sillytavern-runtime-debug