The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs shell commands by directly interpolating user-provided project names into commands like pnpm create vite <项目名称>. This behavior allows for potential command injection if the user input is not properly sanitized or escaped.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) where external user input directly influences executable logic.
Ingestion points: User-provided project names and framework preferences in SKILL.md.
Boundary markers: Absent; there are no instructions for the agent to use delimiters or ignore embedded instructions in the project name.
Capability inventory: The skill has access to shell commands via pnpm, npm, yarn, bun, and npx as defined in SKILL.md.
Sanitization: Absent; no validation or escaping logic is defined to protect against malicious input.
[EXTERNAL_DOWNLOADS]: The skill suggests fetching project templates from external GitHub repositories using the tiged utility. While it mentions well-known community repositories, it also allows arbitrary user/repo combinations.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of code downloaded from external sources through commands like pnpm install and pnpm dev. If a community template is compromised, malicious code could be executed during the project setup process.

vite-starter