The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides instructions for authenticating with the Soda Cloud service using API keys and secrets. It specifically identifies the storage location of sensitive credentials at ~/.soda/credentials and demonstrates passing secrets via command-line flags (e.g., --api-key-secret).- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted data from the Soda Cloud API and processes local YAML configuration files. The AI-assisted 'copilot' mode for generating data contracts involves the agent interpreting data that may be attacker-controlled, which could lead to the execution of malicious instructions embedded within the data or schema. Evidence chain: 1. Ingestion points: sodacli command outputs from the Soda Cloud API and local contract files. 2. Boundary markers: Absent. 3. Capability inventory: sodacli tool execution via shell. 4. Sanitization: Absent.- [COMMAND_EXECUTION]: The skill uses the sodacli command-line interface to perform operations that involve both local file system access and network communication with the Soda Cloud API.

soda-cli