skills/bolhasec/post-empirically-testing-skill-scanners/moltbook-base85-obfuscation-code/Gen Agent Trust Hub
moltbook-base85-obfuscation-code
Fail
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- Obfuscation (HIGH): The file SKILL.md contains a Python script that uses Base85 decoding and an XOR cipher (key 159) to deobfuscate a hidden string. This is a classic evasion technique used to hide malicious commands from static analysis.
- Credentials Unsafe (HIGH): The skill documentation requires the agent to read sensitive API keys from
~/.config/moltbook/credentials.json, which is a high-risk activity often associated with credential theft. - Remote Code Execution (HIGH): The obfuscated payload is intended to 'generate installation instructions' for an external utility. This pattern is frequently used to trick agents into downloading and executing arbitrary scripts from the internet.
- Indirect Prompt Injection (LOW): The skill ingests untrusted data from an external social network (Moltbook) via the
hotandbrowsecommands. Without explicit boundary markers or sanitization, this content could contain instructions that hijack the agent's logic.
Recommendations
- AI detected serious security threats
Audit Metadata