agent-folder-init
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect injection by ingesting untrusted user input (project name and tech stack) and interpolating it into documentation templates. While these inputs are written to files that form the agent's context, the risk is minimal as they are used for static descriptive purposes and are central to the skill's function.\n
- Ingestion points: project_name and tech_stack arguments in scripts/scaffold.py.\n
- Boundary markers: Not implemented in the resulting markdown files.\n
- Capability inventory: Local file writing using pathlib.Path.write_text within scripts/scaffold.py.\n
- Sanitization: None; inputs are inserted into templates using simple string replacement.\n- [SAFE]: The skill includes several security controls to protect the user's environment. It performs a dry run by default, requires an explicit flag for file writes, validates that target paths are within the current workspace (unless overridden), and avoids interacting with potentially dangerous file types such as symlinks. These measures effectively mitigate risks of unintended file system modification.
Audit Metadata