cursor-cancel-ralph
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [DATA_EXPOSURE]: The skill documentation references interacting with paths in the
/root/directory (e.g.,/root/cancel_native_probe). Accessing the root user's home directory is typically restricted and can lead to exposure of sensitive system configuration or logs.\n- [COMMAND_EXECUTION]: The skill performs "native interrupt operations" on processes. This involves process management capabilities that could be misused to terminate unintended tasks if process identification is manipulated.\n- [INDIRECT_PROMPT_INJECTION]: The workflow involves reading loop state files from the filesystem to verify scope and identity.\n - Ingestion points: Workflow step 1 reads "bounded loop state" from local files.\n
- Boundary markers: No explicit delimiters or instructions are specified to ignore embedded instructions.\n
- Capability inventory: The skill can interrupt processes and delete state files.\n
- Sanitization: No evidence of sanitization or validation of the state file content is described.\n- [NO_CODE]: The skill consists of markdown instructions and metadata with no executable scripts or binaries.
Audit Metadata