The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to execute npx pxpipe-proxy, which fetches and runs a third-party package from the public npm registry.
[COMMAND_EXECUTION]: The skill uses shell commands to manipulate ANTHROPIC_BASE_URL, redirecting all agent API traffic through a local proxy. It also uses curl and grep to monitor local system telemetry and proxy logs.
[EXTERNAL_DOWNLOADS]: The skill relies on the download and execution of an external dependency (pxpipe-proxy) that is not hosted by a recognized trusted vendor.
[INDIRECT_PROMPT_INJECTION]: The proxy ingests the complete agent context, including system prompts and session history, to perform image-based compression. There are no explicit boundary markers to prevent the agent from potentially following instructions introduced or modified during the proxy's rendering process. The agent maintains powerful capabilities, including Bash and Write tools, which could be abused if the context is manipulated.
[DATA_EXPOSURE]: The local proxy stores session telemetry and error metadata in ~/.pxpipe/ and provides an unauthenticated local dashboard that serves captured request context.

use-pxpipe