deploy-connector
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several shell commands to manage the connector lifecycle, including running Python wrapper scripts located in the plugin's tools directory and direct calls to the
fivetranCLI for packaging and deployment. - [EXTERNAL_DOWNLOADS]: During the validation step, the
run_connector.pytool downloads and executes a Java-based testing harness. This is a standard component of the Fivetran Connector SDK workflow and is used to verify connector logic before deployment. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading local project files such as
connector.pyandREADME.md. This represents a potential attack surface where malicious instructions could be embedded in code comments or documentation. - Ingestion points: Reads
connector.py,requirements.txt, andREADME.mdfrom the local directory (SKILL.md). - Boundary markers: The skill instructs the agent to look for specific Python code patterns (e.g.,
schema()andupdate()functions) which serves as a structural validation, but it lacks explicit delimiters for the file content when processed. - Capability inventory: The agent has the capability to execute shell commands (
python,fivetranCLI) and perform network operations via the deployment tools. - Sanitization: The skill explicitly forbids the agent from reading, printing, or copying plaintext configuration values into the chat, providing a measure of protection against data exposure.
Audit Metadata