scratchpad
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs routine local file management tasks. No malicious patterns such as credential theft, data exfiltration, or unauthorized network operations were detected.\n- [COMMAND_EXECUTION]: The skill executes standard shell commands (
mkdir -p,ls -la) to create and list the contents of the.scratchpad/directory. These commands are non-privileged and operate within the project's local scope.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential surface for processing untrusted data.\n - Ingestion points: Reads files from the
.scratchpad/directory as described inSKILL.md(edit command).\n - Boundary markers: None identified; instructions do not specify delimiting content read from scratchpad files.\n
- Capability inventory:
mkdir(directory creation),ls(directory listing), and runtime file-editing tools (write/edit/read).\n - Sanitization: Not applicable as the skill relies on the underlying agent's standard file-handling tools.
Audit Metadata