The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill executes npx react-doctor@latest against the user's codebase. Use of the @latest tag ensures that unversioned, dynamic code is downloaded and executed at runtime, which could be replaced by a malicious version without notice.
[REMOTE_CODE_EXECUTION]: The /doctor workflow is designed to fetch a 'remote playbook' from www.react.doctor using curl on every execution. Running remote scripts fetched at runtime is a high-risk pattern as the content of the playbook is outside the user's control and visibility.
[EXTERNAL_DOWNLOADS]: The installation process involves cloning code from a third-party GitHub repository (millionco/react-doctor) using the npx skills add command.
[DYNAMIC_EXECUTION]: The skill relies on fetching and executing logic (playbooks) at use-time, creating a persistent risk of dynamic code injection if the remote server or the package registry is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and scans external React codebases. This creates an attack surface for indirect prompt injection:
Ingestion points: React source files and component code.
Boundary markers: None identified in the provided instructions to separate code data from agent instructions.
Capability inventory: Subprocess execution via npx and network operations via curl.
Sanitization: No evidence of input validation or escaping for the processed code before it influences the agent's analysis logic.
[COMMAND_EXECUTION]: The skill documentation encourages the use of shell commands within the chat interface, which may lead to unintended command execution if the agent is misled by the output of the remote tools.

react-doctor