The Agent Skills Directory

[SAFE]: Detailed auditing of the 65 files in the skill revealed no malicious patterns, command injection vulnerabilities, or unauthorized access attempts.
[DATA_EXFILTRATION]: The skill facilitates communication with a user-configured Tableau Server. This network activity is used exclusively for legitimate API calls (REST, GraphQL, and VDS) to retrieve business data as requested by the user. No connections to untrusted or attacker-controlled domains were found.
[CREDENTIALS_UNSAFE]: Tableau credentials (PATs and passwords) are handled safely using the pydantic-settings pattern, where secrets are loaded from a .env file that is excluded from version control. The data.py module includes a _strip_credentials function that recursively removes sensitive keys from any data structures before they are written to the temp/ folder, preventing accidental credential exposure in logs or exports.
[PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by processing data and metadata from external Tableau environments. However, the instructions in SKILL.md mitigate this risk by requiring the agent to process large datasets in a REPL environment and only surface filtered summaries or small row samples to the context window.
[COMMAND_EXECUTION]: Workflow instructions in SKILL.md advise the agent to use uv run python -c "..." for data exploration. This represents a legitimate use of the agent's code execution capabilities to interact with the provided SDK in a sandbox environment.
[EXTERNAL_DOWNLOADS]: The skill README references standard installation through npx skills add, which is the expected deployment method for the platform and does not involve risky third-party downloads.

query-tableau-data