nx-multi-repo-migrate
Warn
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly suggests disabling the execution sandbox as a workaround for permission errors encountered with the pnpm package manager. Evidence: "Instead, disable the sandbox + restart, or migrate that repo from the unsandboxed parent." (SKILL.md).
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by instructing the agent to execute instructions found in repository data.
- Ingestion points: Files located at tools/ai-migrations/**/*.md and migrations.json (SKILL.md).
- Boundary markers: None. The agent is instructed to "read each prompt and make the described changes".
- Capability inventory: The spawned child agents have access to Bash, Write, and Agent tools, enabling filesystem modification and command execution.
- Sanitization: None. The instructions explicitly command the agent to "apply the deferred prompts" directly.
- [EXTERNAL_DOWNLOADS]: The skill performs package installations and updates from the public npm registry using npm, yarn, pnpm, and bun. Evidence includes calls to nx migrate and package manager install commands targeting external registries.
- [REMOTE_CODE_EXECUTION]: The migration process involves running nx migrate --run-migrations, which executes migration logic and codemods contained within the downloaded packages.
- [CREDENTIALS_UNSAFE]: The skill references sensitive configuration files (e.g., ~/.npmrc, ~/.config/pnpm/rc, ~/.yarnrc.yml) to adjust tool behavior and mentions handling sensitive environment variables like GH_TOKEN to allow for authentication fallback.
Audit Metadata