build-connector

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill interacts exclusively with official Fivetran domains (fivetran.com) for documentation and utilizes the vendor's official SDK (fivetran_connector_sdk). It explicitly instructs the agent to avoid having users paste secrets into chat, recommending secure local alternatives.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from Fivetran's documentation and user-provided descriptions to guide the scaffolding and testing of connectors.
  • Ingestion points: Web content fetched from fivetran.com and user-provided connector descriptions.
  • Boundary markers: No explicit delimiters are specified for the external data ingested from documentation pages.
  • Capability inventory: The skill can execute shell commands for project initialization (fivetran init), dependency installation (uv pip), and testing (run_connector.py).
  • Sanitization: No explicit sanitization or validation of the fetched documentation content is mentioned before it is used to influence build logic.
  • [COMMAND_EXECUTION]: Executes project lifecycle commands including fivetran init and uv. While fivetran init is used with the --yes flag, the instructions mandate that the agent must first verify the directory name and obtain explicit user confirmation before overwriting any existing contents.
  • [EXTERNAL_DOWNLOADS]: Fetches connector catalog data from fivetran.com and installs dependencies from standard package registries. These operations are consistent with the skill's purpose and target the vendor's own resources.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 04:59 AM