The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or unauthorized data exfiltration attempts were identified in the skill instructions or manifest.- [COMMAND_EXECUTION]: The skill utilizes a peloton CLI tool for all operations. The commands are limited to fitness-related tasks such as class searching (search-class), scheduling (schedule-event), and authentication status (status), which are consistent with the skill's stated purpose.- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from the Peloton API, such as class descriptions and titles. While this creates an ingestion surface for untrusted data, the instructions specify structured formatting for output, which mitigates the risk of the agent interpreting data as instructions.
Ingestion points: Results from peloton search-class and browsing commands in SKILL.md.
Boundary markers: Detailed presentation rules for widget.create act as structural boundaries for external content.
Capability inventory: The skill can schedule and delete fitness events but does not have access to sensitive system files or arbitrary network requests.
Sanitization: Data is presented using defined UI components rather than being directly interpolated into the agent's core reasoning path.

peloton