The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution to analyze disk usage and perform deletions, including recursive removal of directories (rm -rf), package management operations (brew uninstall), and container pruning (docker system prune).- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8). It ingests untrusted data from the filesystem (filenames and directory paths) and processes this information to present deletion options to the agent. A malicious file or directory name could contain instructions designed to deceive the agent.
Ingestion points: Output from find, ls, and du commands throughout the scanning phases (e.g., Phase 8: Project Artifacts).
Boundary markers: The skill explicitly mandates using the AskUserQuestion tool for every deletion decision, providing a critical human-in-the-loop verification step.
Capability inventory: Extensive filesystem access and execution of destructive commands like rm -rf, brew uninstall, and xcrun simctl runtime delete.
Sanitization: No explicit sanitization or escaping of the discovered filesystem paths is described before they are interpolated into the agent's context.

disk-cleanup