The Agent Skills Directory

[SAFE]: The skill is designed for administrative planning and incorporates several safety-first principles, including requirements for root authority verification, environment pre-checks, and explicit confirmation before any action is taken. It emphasizes a 'first confirm then execute' approach.- [COMMAND_EXECUTION]: The skill generates plans for system-level changes, such as installing Docker Engine and modifying system repository files. Potential risks are mitigated by requiring that all actions are presented for user review, providing detailed rollback procedures, and instructing the use of backups for critical configuration files.- [EXTERNAL_DOWNLOADS]: An included documentation file (evaluation-report.html) references styling resources from the Tailwind CDN, which is a well-known and trusted service. The skill instructions discuss the use of external shell scripts (e.g., DockerInstallation.sh) but advise against unverified execution and recommend pinning specific versions to avoid supply chain risks.- [DATA_EXPOSURE]: The skill instructions involve reading system-level metadata and configuration files. It specifically cautions against the storage or inclusion of sensitive credentials (tokens, passwords) in the configurations it manages, reducing the risk of accidental credential exposure.- [SAFE]: The skill possesses a surface for indirect prompt injection as it ingests user-provided mirror URLs and reads existing system configuration files. Evidence chain: (1) Ingestion points: SKILL.md arguments and references/operations/preflight.md file reads. (2) Boundary markers: Absent in prompt templates. (3) Capability inventory: High-privilege shell script execution and configuration writes. (4) Sanitization: The skill mandates the use of jq for JSON validation and requires human review of all generated plans before execution.

linuxmirrors-docker