The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the sports-skills package using pip install sports-skills if it is not already present. This package is not pinned to a specific version or a trusted source, which presents a risk of supply-chain attacks or installation of a malicious package if the name were to be hijacked.
[COMMAND_EXECUTION]: The skill heavily relies on the execution of the sports-skills CLI tool and a local shell script (scripts/validate_params.sh) to perform its tasks. While these are intended for data retrieval and parameter validation, they provide a surface for executing shell commands in the agent's environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured and unstructured data from external sources including the ESPN API and the MLB Stats API (statsapi.mlb.com). This data is used to populate parameters for shell commands.
Ingestion points: Data is fetched from statsapi.mlb.com and ESPN public endpoints, including player names, team rosters, and play-by-play descriptions (e.g., event_id, game_pk).
Boundary markers: The scripts/validate_params.sh script attempts to set boundaries by validating date formats (YYYY-MM-DD) and checking for the presence of team IDs before command execution.
Capability inventory: The skill has the capability to execute shell commands (sports-skills), run local bash scripts, and perform network requests to external APIs.
Sanitization: Basic sanitization is implemented via a bash script that uses regex for date validation and basic checks for required identifiers, but it does not comprehensively escape all potential injection characters in complex data fields.

mlb-data