The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The script executes arbitrary commands defined in the target project's package.json file via npm run test, npm run lint, and npm run typecheck. If an AI agent uses this skill on an untrusted repository, the repository can gain full control of the execution environment.
[COMMAND_EXECUTION] (MEDIUM): The script executes shell commands and development tools like find, grep, and npx against the provided project directory. This environment is susceptible to exploitation via malicious project configurations or symlink-based attacks.
[EXTERNAL_DOWNLOADS] (LOW): The script invokes npx tsc, which may trigger the download and execution of the TypeScript compiler from the npm registry if it is not already present in the environment.
[PROMPT_INJECTION] (HIGH): This skill is highly vulnerable to Indirect Prompt Injection (Category 8) because it processes untrusted external data with high-privilege execution capabilities.
Ingestion points: Reads package.json, tsconfig.json, and all source files within the user-provided project directory.
Boundary markers: None. There are no delimiters or instructions to the agent to treat the project data as untrusted or to ignore embedded instructions.
Capability inventory: Possesses the capability to execute arbitrary shell scripts (via npm), run external packages (npx), and perform file system operations.
Sanitization: None. The script trusts and executes the contents of the scripts field in package.json without validation.

refactoring