app-platform-router

Pass

Audited by Gen Agent Trust Hub on May 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the setup of development and troubleshooting environments by fetching resources from external repositories. It instructs the agent to clone the bikramkgupta/do-app-devcontainer repository for devcontainer templates and utilizes container images hosted on ghcr.io/bikramkgupta/ for its sandbox and debug worker components. These downloads provide necessary templates and diagnostic tools used by the skill.
  • [COMMAND_EXECUTION]: Several management scripts, such as add_client.py, cleanup_client.py, and secure_setup.sh, use the Python subprocess module or shell scripts to execute system commands. These commands interface with doctl (DigitalOcean CLI), gh (GitHub CLI), and psql (PostgreSQL client) to automate the provisioning of databases, the configuration of GitHub Secrets, and the management of App Platform components. This behavior is fundamental to the skill's operational tasks.
  • [PROMPT_INJECTION]: The sandbox skill defines a surface for the execution of untrusted code by creating ephemeral, isolated container environments. While the documentation provides examples of eval() and exec() patterns for building AI code interpreters, these are presented as instructional use cases for the sandbox environment rather than executable logic within the skill's own instructions. The skill proactively addresses security by emphasizing that such workflows should occur within the isolated sandbox boundary.
Audit Metadata
Risk Level
SAFE
Analyzed
May 7, 2026, 06:20 PM