The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands using npx to install packages globally (-g). It performs environment checks for Trae-specific directories (~/.trae or ~/.trae-cn) to modify command arguments at runtime.
[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and install external packages from the NPM registry, which are not restricted to a pre-defined whitelist.
[REMOTE_CODE_EXECUTION]: The command npx skills add <package> involves downloading and executing code from the NPM registry. Since the package name is dynamically provided by the find-skills dependency based on user intent, it presents a path for executing arbitrary remote code.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests data (package names) from the output of the find-skills tool.
Boundary markers: Absent. No delimiters or instructions are used to prevent the agent from executing potentially malicious package names returned by the discovery process.
Capability inventory: The skill has the capability to execute shell commands with global installation privileges.
Sanitization: There is no evidence of validation or sanitization of the <package> variable before it is interpolated into the shell command string.

trae-skill-finder