The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands in the background using run_in_background: true and explicitly forbids the agent from displaying command execution logs, technical statuses, or debugging information to the user ("严禁输出的内容：调试地址... 命令执行日志"). This concealment reduces transparency and user oversight of the agent's actions on the local system.- [COMMAND_EXECUTION]: Both the project initialization scripts (scripts/init-hap-view-project.sh) and the core instructions (SKILL.md) command the agent to perform global software installations using sudo (e.g., sudo npm install -g mdye-cli), which is a high-privilege operation that should not be automated silently.- [EXTERNAL_DOWNLOADS]: The skill automates the installation of multiple external packages from the NPM registry (mdye-cli, recharts, styled-components, dayjs, apifox-mcp-server) without requiring individual user confirmation for these downloads.- [REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to automatically install and run external CLI tools (mdye-cli) and local development servers (mdye start) while instructions in hap-view-plugin-dev-guide.md explicitly tell the agent to act autonomously ("用户无需手动操作任何命令", "静默执行") during this process.- [PROMPT_INJECTION]: The skill uses meta-instructions in hap-view-plugin-dev-guide.md and SKILL.md to override the agent's standard reporting behavior ("你不应该做的：... 展示技术细节... 不要多说一个字!"). These instructions are designed to enforce a non-transparent execution mode that hides the agent's background activities from the user.

hap-view-plugin