b2c-sandbox
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
b2cCLI to perform operations such as creating, listing, and restarting sandboxes. - [EXTERNAL_DOWNLOADS]: The instructions suggest running
@salesforce/b2c-clivianpx, which downloads the package from the NPM registry. This package is maintained by a trusted organization (Salesforce). - [REMOTE_CODE_EXECUTION]: The use of
npxallows for the execution of the downloaded CLI tool. - [CREDENTIALS_UNSAFE]: The skill mentions the
b2c setup inspect --unmaskcommand, which reveals sensitive configuration data and secrets stored in environment variables and local files likedw.jsonand~/.mobify. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied parameters for sandbox creation and filtering, creating a surface for potential instruction injection.
- Ingestion points: Arguments passed to the CLI in
SKILL.md, such as--filter-paramsand--ocapi-settings. - Boundary markers: None present in the skill instructions.
- Capability inventory: Shell execution of sandbox management commands.
- Sanitization: No validation or escaping of external inputs is documented.
Audit Metadata