Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 94% confidence
- Finding
- The skill is presented as a Tolstoy integration, but the documented setup performs a local configuration change by writing to `~/.openclaw/openclaw.json` or an alternate path from an environment variable. That side effect broadens trust to a remote MCP endpoint and can alter future agent behavior persistently, yet it is not surfaced as a sensitive operation with clear security implications.
