The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill triggers the download of the react-doctor package from the public npm registry at runtime using the npx command.
[REMOTE_CODE_EXECUTION]: The use of npx -y react-doctor@latest allows for the execution of remote code. By specifying the @latest tag, the skill does not pin a specific version, meaning the behavior of the skill can change if the remote package is updated or compromised.
[COMMAND_EXECUTION]: The skill executes a shell command to perform its primary function. This command runs with the permissions of the agent and accesses the project's root directory (.).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes and acts upon untrusted data from the codebase being scanned.
Ingestion points: The agent reads diagnostics produced by the react-doctor tool, which may include fragments or metadata from the scanned source files.
Boundary markers: There are no explicit delimiters or instructions provided to help the agent distinguish between legitimate diagnostic messages and potential malicious instructions embedded in the scanned code comments or strings.
Capability inventory: The agent is instructed to modify the codebase ("Fix issues") based on the content of these diagnostics.
Sanitization: The skill does not implement any validation or sanitization of the tool's output before it is presented to the agent for action.

react-doctor