The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses the npm install command to download and execute code from the npm registry, which is a well-known service for Node.js packages.
[COMMAND_EXECUTION]: The skill enables the execution of arbitrary scripts defined in the project's configuration through the npm run command.
[DYNAMIC_EXECUTION]: Package installation via npm install often triggers the execution of lifecycle scripts (e.g., preinstall, postinstall) provided by package authors.
[DATA_EXFILTRATION]: The npm publish action uploads local directory contents to a remote registry, which could lead to the unintended exposure of sensitive local files or configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes project configuration files that could contain malicious instructions.
Ingestion points: Instructions direct the agent to read and verify the package.json file (SKILL.md).
Boundary markers: There are no specific instructions to treat content from the configuration files as untrusted data or to use delimiters.
Capability inventory: The skill has the ability to install external code, execute local shell commands, and publish data to external registries.
Sanitization: No sanitization or verification of the package.json content is mentioned before it is processed by the agent.

npm