skill-creator
Warn
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs shell-level orchestration, including running background processes with 'nohup' and process termination using 'kill'. It also executes several Python modules directly via command-line interfaces.
- [REMOTE_CODE_EXECUTION]: The workflow involves the dynamic generation and execution of Python scripts (e.g., 'create_docx.py') based on task results. It also depends on several internal scripts ('aggregate_benchmark', 'generate_review.py', 'run_loop.py') that are not provided in the source distribution.
- [DATA_EXFILTRATION]: The skill is instructed to read from the '~/Downloads' directory to import datasets, creating a risk of unintended access to sensitive user files.
- [PROMPT_INJECTION]: The skill provides instructions for manipulating model triggering and includes an indirect prompt injection surface. Evidence Chain: 1. Ingestion points: Reads 'feedback.json' and 'evals.json' containing user-controlled text. 2. Boundary markers: None identified. 3. Capability inventory: Shell execution, script generation, and subagent spawning. 4. Sanitization: No validation or filtering of user-provided content is performed.
Audit Metadata