The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains explicit instructions for the agent to automatically trigger an 'auto run' build process involving shell commands. Specifically, it directs the execution of ohpm install and hvigorw assembleHap when structural code changes or dependency modifications occur. Automated execution of package managers and build tools is dangerous because these tools often execute arbitrary scripts defined within the project (e.g., pre-install or build scripts), providing a path for local code execution.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a significant attack surface by combining the ingestion of untrusted user code with the capability to execute system-level commands. Ingestion points: The agent processes user-provided .ets source files and configuration files such as oh-package.json5 and module.json5. Boundary markers: There are no instructions provided to help the agent distinguish between its own instructions and potentially malicious instructions embedded in the code it analyzes. Capability inventory: The skill explicitly grants the agent the capability to run shell commands like ohpm and hvigorw. Sanitization: The skill does not provide any mechanisms for sanitizing project configurations or validating scripts before they are executed.
[PROMPT_INJECTION]: The skill employs strong imperative language ('强制要求', '必须严格遵守') to override standard agent behavior. It specifically instructs the agent to prioritize automated build flows over manual verification and 'forbids' (禁止) relying solely on static analysis when structural changes are detected.

harmony-arkts-development-assistant