The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to install an external package named sports-skills. It provides a direct installation path from a GitHub repository: https://github.com/machina-sports/sports-skills.git. This package is provided by the skill's author and is required for functionality.
[COMMAND_EXECUTION]: The skill uses local shell commands to check for the presence of the sports-skills CLI and includes a helper script scripts/validate_params.sh that uses shell-based parameter validation logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (TFRRS and The Stride Report), creating a surface for potential indirect prompt injection attacks where malicious instructions could be embedded in sports data.
Ingestion points: Data is retrieved from athlete profiles (names, PRs) and news article summaries via external URL requests in the get_athlete_profile, get_meet_results, and get_news commands.
Boundary markers: No specific delimiters or instructions to the agent are defined to separate user/agent instructions from the fetched athlete or news data.
Capability inventory: The skill leverages the sports-skills CLI to make network requests to tfrrs.org and thestridereport.com.
Sanitization: The instructions do not define sanitization procedures for the external content before it is processed by the agent.

xctf-data