soia-env-storage-cleanup
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local Python script
scripts/storage_cleanup.pyto perform file system scanning, plan generation, and authorized file deletion. - [DATA_EXPOSURE]: The script accesses file system metadata, including paths, sizes, and timestamps, within directories explicitly defined by environment variables (e.g.,
SOIA_SKILLS_CACHE_HOME) to identify cleanup candidates. - [SAFE]: Implements a strict two-step authorization workflow where a read-only plan is first presented to the user, requiring the user to explicitly confirm a unique
plan_idbefore any deletion occurs. - [SAFE]: Uses SHA-256 digests to verify the integrity of cleanup plans and receipts, ensuring that the files targeted for deletion have not been tampered with between the planning and execution phases.
- [SAFE]: Enforces directory boundaries using mandatory marker files (
.soia-storage-root.json) and validates that all candidates are regular files (not symlinks) located within the managed roots. - [SAFE]: Includes "fail-closed" logic that automatically skips files if their size or modification time changes after the plan is generated, or if they are marked as active by the system.
Audit Metadata