The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install the peloton-pp-cli binary using the npx package runner and the go install command. These downloads originate from the npm registry and GitHub repositories associated with the developer 'mvanhorn'.
[REMOTE_CODE_EXECUTION]: The installation process involves executing remote code from package registries (via npx) and compiling source code from a remote repository (via go install). While these are standard developer workflows, they represent a point where external code is introduced into the environment.
[COMMAND_EXECUTION]: The skill's primary function is to execute the peloton-pp-cli tool to interact with Peloton's data services and local stores.
[DATA_EXFILTRATION]: The CLI tool contains a built-in feature (--deliver webhook:<url>) that allows the agent to POST command results to an arbitrary remote URL. It also includes a feedback command that can be configured via environment variables to send local data to an external server.
[PROMPT_INJECTION]: The skill processes data from the Peloton API, creating an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent context through Peloton workout history, class catalogs, and account profiles retrieved via the CLI.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to distinguish between the tool's output and potentially malicious data embedded within the API responses.
Capability inventory: The skill has the ability to write to files (--deliver file:) and perform network requests (--deliver webhook:, feedback --send).
Sanitization: The skill does not mention any sanitization, filtering, or validation of the data received from external API endpoints before it is processed by the agent.

pp-peloton