relay-dispatch

Pass

Audited by Gen Agent Trust Hub on May 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill orchestrates workflows by invoking system utilities such as git, gh (GitHub CLI), and sqlite3. These are used for worktree creation, pull request automation, and maintaining local execution state. The scripts validate paths using fs.realpathSync and containment checks to ensure commands operate within authorized directories.
  • [REMOTE_CODE_EXECUTION]: The skill dispatches task prompts to external AI executors (claude, codex, opencode) which can perform implementation work in the repo. The Claude executor explicitly uses the --dangerously-skip-permissions flag to enable automated implementation without manual approval for every tool call. This is a design choice for non-interactive automation, and users are warned when executors (like opencode) lack native sandboxing.
  • [EXTERNAL_DOWNLOADS]: During the dispatch process, the skill performs git fetch operations to synchronize isolated worktrees with the remote repository's base branch. This ensures implementation tasks are executed against current code state, preventing conflicts during the eventual merge phase.
Audit Metadata
Risk Level
SAFE
Analyzed
May 24, 2026, 03:00 AM
Security Audit — agent-trust-hub — relay-dispatch