okp-import
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions specify the installation of the
@markbangwu/okppackage from the npm registry as a prerequisite for operation. - Evidence:
npm install -g @markbangwu/okpinSKILL.md. - [COMMAND_EXECUTION]: The skill workflow relies heavily on the execution of the
okpCLI tool to manage domains, handle invitations, and perform data import operations. - Evidence: Multiple bash command examples including
okp domain <domain> --set readme.md,okp invite accept,okp put, andokp batchinSKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data (domain README files and concept content) to "distill" structured JSON. This ingestion of untrusted content represents a vulnerability surface where embedded instructions could attempt to influence the agent's behavior.
- Ingestion points:
okp domain <domain>(retrieves remote README content) and Step 3 (distillation of external raw data) inSKILL.md. - Boundary markers: None specified in the distillation instructions.
- Capability inventory: Network communication with the OKP API and command execution via the CLI tool.
- Sanitization: While Step 4 mentions schema validation for fields, there is no evidence of prompt-level sanitization for ingested text content.
Audit Metadata