setup-project

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads project templates from the author's GitHub repository (gh:RobertDeRose/dstack). It also orchestrates the download and installation of various standard developer tools via the mise manager, including uv, mdbook, ruff, and hk (v1.49.0).
  • [COMMAND_EXECUTION]: The skill executes shell commands for Git repository initialization, mise tool management, and bd (Beads) workflow initialization. These operations are performed using subprocess.run with structured argument lists, mitigating command injection risks.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data (such as project purpose, scope, and boundaries) collected via AskUserQuestion and interpolates them into Markdown documentation templates (AGENTS.md, README.md).
  • Ingestion points: User input from AskUserQuestion in SKILL.md.
  • Boundary markers: Templates use explicit <!-- BEGIN DSTACK WORKFLOW --> comments to delimit generated content.
  • Capability inventory: Writing to the filesystem and executing shell commands (Git, mise).
  • Sanitization: The skill employs a Jinja2 macro (markdown) in template files to escape Markdown special characters (e.g., #, *, [, ], `), preventing user-provided text from breaking the documentation structure or injecting malicious Markdown instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 11:29 AM
Security Audit — agent-trust-hub — setup-project