parallel-dev-worktrees
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill manages developer workflows by executing Git commands and package manager scripts (npm, pnpm, yarn, bun) to manage worktrees and dependencies.
- [COMMAND_EXECUTION]: It prioritizes and executes repository-defined scripts (e.g.,
wt:create,wt:doctor) for project-specific automation. - [EXTERNAL_DOWNLOADS]: Fetches and installs the companion Portless skill from the Vercel Labs official GitHub repository, which is a well-known and trusted source.
- [SAFE]: Implements several security-best practices, such as preventing operations on dirty worktrees, isolating local environment variables, and ensuring clean integration branch checkouts before merging.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes untrusted project documentation and scripts. (1) Ingestion points: Reads project-level files such as
AGENTS.md,package.json, and documentation inwiki/plans/. (2) Boundary markers: No explicit boundary markers or delimiters are specified for ingested data. (3) Capability inventory: Includes shell command execution via git and package managers, file system writes, and browser automation. (4) Sanitization: Relies on the agent's default processing and manual user oversight; no specific sanitization of project-provided scripts or plan data is mentioned.
Audit Metadata