visualization
Pass
Audited by Gen Agent Trust Hub on Apr 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command to activate a Python virtual environment using a hardcoded absolute path. Evidence:
source /Users/zhangmingda/clawd/.venv/bin/activatein SKILL.md. - [DATA_EXFILTRATION]: Hardcoded absolute file paths reveal a local system username. Evidence:
/Users/zhangmingda/path in SKILL.md. - [PROMPT_INJECTION]: The skill processes untrusted data frames to generate visualizations without sanitization or boundary markers, creating an indirect prompt injection surface. Evidence: 1. Ingestion points: Dataframes (
df) used in various plotting examples in SKILL.md. 2. Boundary markers: Absent. No instructions are provided to ignore embedded instructions or delimit the data. 3. Capability inventory: Python execution via standard plotting libraries for file writing and image generation. 4. Sanitization: Absent. The data is passed directly to third-party visualization libraries.
Audit Metadata