coupling-router
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a local Bun script,
scripts/worktree-lease.ts, to coordinate shared git worktrees between multiple agent sessions. This script executesgitcommands (viaBun.spawnSync) to check branch status and manage lock files, providing a coordination mechanism for multi-session agent environments without external network activity. - [INDIRECT_PROMPT_INJECTION]: As an orchestration skill, it processes and analyzes task lists, external skill instructions, and file diffs. This ingestion of external data constitutes an attack surface for indirect prompt injection. The skill mitigates this through a mandatory 'Plan-Evaluation Gate' which requires verifiable acceptance criteria, spec alignment, and evidence-backed assumptions for all tasks, providing a validation layer for processed inputs.
Audit Metadata