openclaw-proof-validation
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to execute various local scripts and binaries, such as
build.ps1,run-app-local.ps1, andparallels-windows-vm.sh. These tools are used to build the software, run automated tests, and manage virtual machine snapshots for validation. - [DATA_EXFILTRATION]: The instructions provide methods for the agent to retrieve
mcp-token.txtfrom theOpenClawTrayapplication data directory. This token is used to authenticate to the local Model Context Protocol (MCP) server for tool discovery and invocation during testing; no external exfiltration is instructed. - [PRIVILEGE_ESCALATION]: The remote execution guidelines for the Parallels VM backend include instructions to use the SYSTEM account specifically for machine installers. This represents a requirement for high-privilege access during system-level software setup within the isolated VM environment.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and reporting on output from external sources, including test results, UI diagnostics, and JSON-RPC responses from the local MCP server. This data ingestion creates a surface where malicious content within those outputs could theoretically influence agent behavior, although the skill includes process rules for objective reporting.
Audit Metadata