project-bootstrap
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and installs project-local utilities from the author's repository using the command 'npx skills@latest add yikZero/skills --skill find-docs' (referenced in SKILL.md and references/skill-recommendation.md).
- [REMOTE_CODE_EXECUTION]: By invoking 'npx skills' and 'npx yskill', the skill executes code from remote npm packages during the initialization process.
- [COMMAND_EXECUTION]: The skill executes a local TypeScript script 'scripts/bootstrap-project.ts' using the Bun runtime to generate project files and directories based on user-selected profiles.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8). It ingests untrusted user input (such as project names and descriptions) and discovery results, then interpolates this data into generated control documents (README.md, AGENTS.md, INIT.md) which guide future agent behavior. Ingestion points: Command-line arguments in 'scripts/bootstrap-project.ts' and search results in the 'Best-Practice Discovery' workflow. Boundary markers: The generated templates lack explicit delimiters or instructions to ignore embedded commands in the interpolated fields. Capability inventory: The skill script has filesystem write access. The generated project configuration (.codex/config.toml) defaults to high-privilege 'sandbox_mode = "danger-full-access"' and enables network access. Sanitization: User-provided strings are placed directly into template files without escaping or validation.
Audit Metadata