lov-wdb-cli
Fail
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: HIGHOBFUSCATIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [OBFUSCATION]: The skill employs total encryption of its operational logic. The actual instructions and Python scripts are stored as encrypted blobs (e.g., SKILL.md.enc, scripts/wdb_cli.py.enc). The visible SKILL.md acts as a 'loader', instructing the AI agent to decrypt the payload at runtime using an external tool. This technique intentionally evades static analysis and prevents users or automated scanners from auditing the skill's actual behavior.
- [REMOTE_CODE_EXECUTION]: The skill requires the execution of remote packages at runtime. It instructs the agent to use 'uvx lovstudio-skill-helper' and 'npx lovstudio'. These tools are fetched from public registries and executed, presenting a significant supply-chain risk, especially as they are used to handle sensitive decryption keys and personal data.
- [COMMAND_EXECUTION]: The skill relies on the execution of multiple shell commands to decrypt its files and manage licenses.
- [DATA_EXFILTRATION]: While direct exfiltration isn't visible in the loader, the skill's stated purpose is to access highly sensitive local data, including WeChat chat history, contact lists, and social media feeds ('moments'). The combination of sensitive data access with an un-auditable, encrypted codebase represents a severe privacy risk.
Recommendations
- AI detected serious security threats
Audit Metadata