The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/expand-type.ts uses module.createRequire to dynamically load the typescript module from a path computed from the target file's location. This allows for the execution of arbitrary code if the analyzed directory contains a malicious package.json or node_modules folder.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using pnpm exec tsx. This executes a local script provided with the skill, utilizing the system's shell environment and allowing for execution of the script at a local path.
[EXTERNAL_DOWNLOADS]: The skill documentation refers to pnpm dlx tsx, which involves fetching the tsx package from the public npm registry for execution.
[PROMPT_INJECTION]: The skill reads and processes the content of external TypeScript files, which serves as an indirect prompt injection surface. Maliciously crafted source code or comments could potentially influence the agent's behavior.
Ingestion points: Source files specified via the --file argument.
Boundary markers: Expanded types are returned within markdown code blocks.
Capability inventory: The skill has file system read access and shell command execution capabilities.
Sanitization: The script uses the TypeScript compiler API to parse types, but does not explicitly filter or sanitize comments within the ingested files.

type-expand