design-review
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads and executes an installation script for the Bun runtime. It implements a critical security check by validating the script's SHA256 checksum (bab8acfb046aac8c72407bdcce903957665d655d7acaa3e11c7c4616beae68dd) against a hardcoded value before execution.\n- [EXTERNAL_DOWNLOADS]: Fetches the Bun installer from https://bun.sh, which is a well-known development service.\n- [COMMAND_EXECUTION]: Executes shell commands to setup the environment, install project-specific packages (npm, pip, etc.), and perform git operations. The skill includes human-in-the-loop safeguards, using interactive prompts to seek user approval before committing changes or modifying test configurations.\n- [PROMPT_INJECTION]: The skill processes content from external web pages and project source code to identify design issues, creating an attack surface for indirect prompt injection.\n
- Ingestion points: Ingests data from external websites via browser tools and local source files via Read/Glob tools.\n
- Boundary markers: No explicit boundary markers or isolation instructions are used when processing audited content.\n
- Capability inventory: Includes file modification (Edit/Write), shell command execution (Bash), and network/browser access.\n
- Sanitization: No specific sanitization or filtering of ingested audit data is defined.
Audit Metadata