setup-project
Fail
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill automates the installation of development tools using shell pipes from remote sources. Specifically, it executes installation scripts for
uv(from Astral),bun,cargo(Rust), and.NET(from Microsoft). - [EXTERNAL_DOWNLOADS]: Fetches installation scripts and project metadata from official and well-known domains, including
astral.sh,bun.com,sh.rustup.rs, anddot.net. It also retrieves versioning information and skill definitions from the author's repository on GitHub (raw.githubusercontent.com/Gharib89/skills). - [INDIRECT_PROMPT_INJECTION]: The skill contains a mechanism to "follow" instructions from external markdown files (
SKILL.md) located in the.claude/skills/directory. These instructions are treated as if they were part of the primary skill's execution flow. - Ingestion points: Files located at
.claude/skills/<name>/SKILL.mdare read and their processes are executed. - Boundary markers: There are no explicit delimiters or warnings to ignore potentially malicious embedded instructions in the external files.
- Capability inventory: The agent is granted capabilities to perform shell command execution, git repository management, and interaction with various CLI tools (
gh,az,npx). - Sanitization: The content of the external skill files is not validated or sanitized before execution.
- [COMMAND_EXECUTION]: The skill performs extensive command-line operations to manage the development environment. This includes initializing git repositories, creating GitHub repositories via
gh repo create, and managing local project structures.
Recommendations
- HIGH: Downloads and executes remote code from: https://astral.sh/uv/install.sh, https://bun.com/install, https://sh.rustup.rs - DO NOT USE without thorough review
Audit Metadata