nuke-release
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to verify installation and usage claims by executing commands directly from the repository's documentation (README, quickstart guides, etc.). This capability allows the skill to run arbitrary shell commands present in the files it analyzes.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from documentation and manifests to drive its verification pipeline, creating an injection surface.
- Ingestion points: README.md, quickstart guides, and general documentation files are parsed for executable truth claims (SKILL.md, Phase 1 Charter 1).
- Boundary markers: The skill explicitly mandates tracing destructive or publishing commands (deploy, publish, tag) rather than executing them (Mandate 2), which provides a limited safety boundary.
- Capability inventory: The skill uses subprocess execution to run installation steps, requirement statements, and usage examples (SKILL.md, Mandate 2).
- Sanitization: No explicit sanitization or filtering of documentation content is mentioned before the extracted commands are passed to the shell.
- [EXTERNAL_DOWNLOADS]: The preflight protocol allows the skill to resolve environmental blockers by automatically installing missing dependencies using standard package managers (npm, pip, go, etc.) if the documented commands fail due to a missing environment (preflight.md, Blocked preflight).
Audit Metadata