harmonyos
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents a 'vision-driven device automation' workflow in
references/build-sign-verify.mdandSKILL.mdthat leverages natural language instructions to perform device interactions, creating a potential vector for indirect prompt injection. - Ingestion points: Natural language descriptions of interactions are processed by automation tools as described in
references/build-sign-verify.md. - Boundary markers: No specific delimiters or safety warnings for the natural language automation loop are provided.
- Capability inventory: The automation tools possess capabilities to execute
hdc shellcommands and perform UI events like clicks and text input via theuitestsystem utility. - Sanitization: The documentation does not specify validation or sanitization of the natural language instructions before they are processed by the automation model.
- [EXTERNAL_DOWNLOADS]: The skill instructs users to use
npx -y @midscene/harmony@1within thereferences/build-sign-verify.mdfile, which involves downloading and executing code from the NPM registry to facilitate device automation. - [COMMAND_EXECUTION]: The skill involves the execution of various command-line tools such as
hvigorw,ohpm,hdc, anduitestfor building applications, managing dependencies, and verifying behavior on connected HarmonyOS devices.
Audit Metadata