connect-hubspot-agent-cli

Fail

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the HubSpot Agent CLI installation script and a PowerShell equivalent from HubSpot's official API domain (api.hubapi.com). It also uses npx to fetch additional agent skills from the HubSpot organization.
  • [REMOTE_CODE_EXECUTION]: The installation process involves piping a remote script directly to a shell interpreter (sh on Linux/macOS or iex on Windows). This pattern is a common distribution method for CLI tools from well-known technology vendors.
  • [COMMAND_EXECUTION]: The skill relies on shell command execution to perform HubSpot administrative tasks, including record deletion, workflow management, and schema updates. It provides specific instructions for the agent to use dry-run flags and verify the target portal using hubspot whoami before executing mutations.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: The skill ingests data from HubSpot's CRM (objects, pipelines, and workflows) using search and list commands, which could contain untrusted content.
  • Boundary markers: The skill includes explicit instructions for the agent to perform dry runs, show previews to the user, and wait for confirmation before any write operation.
  • Capability inventory: The skill allows the agent to execute the hubspot CLI for read/write/delete operations and permits writing data to the local file system (e.g., data/audit-logs/).
  • Sanitization: The instructions state that generated JSONL should be treated as untrusted until reviewed and recommend sampled validation of bulk data.
Recommendations
  • HIGH: Downloads and executes remote code from: https://api.hubapi.com/hub/cli/backend/hub-cli/latest/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 16, 2026, 11:09 PM