linux-cloud-init
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions and documentation recommend downloading configuration files and bootstrap scripts from the author's GitHub repository (github.com/petebwire/linux-skills.git). It also references official repositories for well-known services including Docker (download.docker.com) and NodeSource (deb.nodesource.com).
- [REMOTE_CODE_EXECUTION]: Documentation and worked examples provide instructions for the agent to download and execute shell scripts (e.g., setup-claude-code.sh and install-skills-bin) with elevated privileges (sudo bash) on the systems being provisioned.
- [COMMAND_EXECUTION]: The skill workflow involves executing shell commands to validate, debug, and test configurations, including the use of 'lxc exec' to interact with test containers and 'cloud-init' system tools to analyze boot states.
- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection through the processing of untrusted data.
- Ingestion points: The skill is designed to ingest and act upon user-supplied 'user-data' and 'autoinstall' YAML files (as described in SKILL.md and references/debugging.md).
- Boundary markers: There are no explicit boundary markers or delimiters configured to prevent the agent from interpreting instructions embedded within the processed configuration data.
- Capability inventory: The skill utilizes powerful capabilities including file system writes (write_files), package installations, and arbitrary command execution (runcmd) at the system level.
- Sanitization: While the skill recommends the use of 'cloud-init schema' to validate YAML syntax and structure, it lacks specific mechanisms to sanitize or filter natural language instructions embedded within the data values.
Audit Metadata