scaffolding-maintainer
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes 'assets/observations.md', an append-only log of user feedback and rejections, to drive improvements to the scaffolder's templates and logic. This ingestion of unsanitized external data represents a potential injection surface.
- Ingestion points: 'assets/observations.md' is read by the agent to determine necessary changes.
- Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the log entries.
- Capability inventory: The agent has the ability to execute shell commands (Python, npm, git, zip) and modify the skill's own source code and configuration files.
- Sanitization: The skill relies on evidence thresholds rather than input sanitization.
- [COMMAND_EXECUTION]: The skill relies on shell command execution for verification and maintenance tasks.
- 'scripts/verify_scaffold.py' executes 'subprocess.run' to call 'scaffold.py' and run 'npm install' and 'npm test' suites on generated projects.
- 'SKILL.md' contains instructions for the agent to use 'zip' for repackaging and 'git' for baseline management.
- [EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch registry data and install dependencies from well-known public registries.
- 'scripts/verify_scaffold.py' runs 'npm install', which fetches packages from the public npm registry.
- The 'refresh_versions.py' script queries npm and PyPI registries to check for package updates.
Audit Metadata