cleanup
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/agent_cleanup.pyexecutes various system utilities (e.g.,git,npm,docker,du) to manage repositories and package caches. These are executed using list-based arguments insubprocess.run, which prevents shell command injection.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data such as git branch names and file system paths which are subsequently displayed to the AI agent for confirmation. A malicious repository could use these names to attempt to influence the agent's behavior.\n - Ingestion points: Local branch names and file paths read via
gitandos.walkinscripts/agent_cleanup.py.\n - Boundary markers: The instructions do not define specific delimiters for separating untrusted branch/path strings from agent instructions.\n
- Capability inventory: The skill can perform destructive file system operations, including directory deletion (
shutil.rmtree) and git branch removal.\n - Sanitization: There is no evidence of sanitization or escaping of the branch names and file paths before they are output to the agent context.
Audit Metadata