The Agent Skills Directory

File System Operations: The scripts scripts/opentargets_graphql.py and scripts/opentargets_disease_heatmap.py allow the agent to specify local paths for reading queries and writing output. The query_path parameter enables reading from the local file system, while the raw_output_path parameter allows the scripts to create directories and write files. These features represent a potential risk if the agent is instructed to access or overwrite sensitive system files using these parameters.
Network Communication: The skill performs network requests to api.platform.opentargets.org to retrieve data. This is a well-known and expected service for the skill's purpose, and the communication is conducted over HTTPS.
Data Processing Surface: The skill ingests data from a remote GraphQL API. Although the scripts use truncation to limit the volume of data returned to the agent, the combination of processing external data and maintaining file system access creates a surface for potential indirect influence if the external data source were to provide unexpected content.
Input Validation: The scripts include logic to validate the structure and types of the input JSON payloads, which helps ensure that the tools are used as intended and reduces the risk of malformed input causing issues.

opentargets-skill