The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with OS-level utilities like ps, lsof, taskkill, handle.exe, and openfiles. These calls are used to detect processes that are locking files and to monitor disk usage. This is standard behavior for a disk management tool and is implemented using safe practices (list-based arguments without shell interpretation).
[DYNAMIC_EXECUTION]: The skill_bootstrap.py module uses __import__ to verify the availability of its core modules in different environments. Additionally, check_skill.py uses compile() to verify the syntax of the package's scripts as part of a diagnostic test. Both uses are restricted to the local package context and are functional in nature.
[INDIRECT_PROMPT_INJECTION]: The skill presents file and directory names from the local filesystem to the agent, creating a data ingestion surface.
Ingestion points: diskcleaner/core/scanner.py reads directory entries and metadata.
Boundary markers: SKILL.md defines strict ASCII-safe output markers (e.g., [OK], [!]) for scripts and provides templates for agent-to-human reporting.
Capability inventory: The skill possesses file deletion capabilities (scripts/clean_disk.py) and process termination capabilities (diskcleaner/core/process_manager.py).
Sanitization: While filenames are processed as raw strings, the skill enforces safety through dry-run defaults and requires explicit user confirmation (e.g., typing 'YES' or 'I UNDERSTAND') before performing destructive operations.

disk-cleaner