setup-application
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
branchly_prefixed tools (e.g.,branchly_get_application,branchly_update_data_source,branchly_create_prompt) through a connected MCP server. These commands are scoped to the management of the branchly application and do not show signs of arbitrary shell execution or privilege escalation. - [EXTERNAL_DOWNLOADS]: The skill references official documentation and dashboards at
docs.branchly.ioanddashboard.branchly.io. These are vendor-owned resources and are used for providing context and tracking progress, which is standard for a setup utility. - [INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection in Phase 3 and Phase 7, where it processes external website content via crawlers and external data via the Public QA API. However, it implements risk-reduction measures such as explicit user consent for QA testing, noise-stripping CSS selectors (
remove_html_elements) to filter out untrusted UI elements, and instructions for the agent to maintain strict scope discipline and avoid duplicating system-level safety rules.
Audit Metadata