lov-wdb-cli
Fail
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [OBFUSCATION]: The skill's operational logic, including its primary instructions and Python scripts, is distributed in an encrypted format using AES-256-GCM. This prevents static analysis and hides the skill's behavior from security review.
- Evidence: Encrypted files SKILL.md.enc, references/data-contract.md.enc, and scripts/*.py.enc.
- [DATA_EXFILTRATION]: The skill is explicitly designed to search for and extract WeChat database management keys and read private local chat data. Accessing third-party encrypted messaging databases is a highly sensitive operation that can lead to significant privacy breaches.
- Evidence: Description in SKILL.md stating the ability to "automatically discover WDB Pro management keys" and "precisely read local WeChat data."
- [REMOTE_CODE_EXECUTION]: The skill relies on external command-line tools (npx lovstudio, uvx lovstudio-skill-helper) to download, decrypt, and facilitate the execution of the skill's logic. This introduces unverified external code into the execution environment at runtime.
- Evidence: Installation and decryption commands provided in SKILL.md.
- [INDIRECT_PROMPT_INJECTION]: The skill processes large volumes of external, untrusted data (local WeChat logs). This represents an attack surface where instructions embedded within chat messages could influence the agent's behavior.
- Ingestion points: WeChat local databases (referenced in scripts/wdb_cli.py.enc logic).
- Boundary markers: Verification impossible due to file encryption.
- Capability inventory: Reading sensitive local files and discovering system-level keys.
- Sanitization: Verification impossible due to file encryption.
Recommendations
- AI detected serious security threats
Audit Metadata