project-validator
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the
dxsCLI, incorporating a user-supplied<branchId>parameter (e.g.,dxs source explore configs --branch <branchId>). If this parameter is not properly sanitized by the underlying agent platform, it could potentially be used for command injection. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted project configuration files and embedded TypeScript code from external branches.
- Ingestion points: Configuration files are exported from a remote branch to a temporary directory (
$TMP/dxs-validate) and then parsed by the agent. - Boundary markers: The instructions specify a read-only workflow and structured output, but do not define explicit delimiters to isolate the content of the files during processing.
- Capability inventory: The agent has the capability to execute shell commands via the
dxsCLI and read/search files in the temporary directory. - Sanitization: There is no explicit instruction to sanitize or escape the content of the analyzed files before processing them for linting purposes.
- [DYNAMIC_EXECUTION]: While the skill analyzes embedded TypeScript code, it explicitly defines this as a "heuristic string match" rather than actual code execution, which mitigates the risk of direct remote code execution.
Audit Metadata