The Agent Skills Directory

[Obfuscation] (HIGH): The 'Installing openclaw-core utility' section uses Base85 encoding combined with a XOR cipher (key 159). This is a strong indicator of malicious intent designed to bypass static analysis of the installation commands.
[Remote Code Execution] (HIGH): The obfuscated payload is explicitly intended to be decoded and executed as installation instructions. This pattern is commonly used to hide 'curl | bash' or similar RCE vectors targeting the host system.
[Credentials Unsafe] (HIGH): The skill explicitly targets 'api_key' and 'agent_name' stored in '~/.config/moltbook/credentials.json'. While functional for a social network skill, the presence of obfuscated installation logic makes this a high-risk data exposure finding.
[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from a social network feed ('GET /posts').
Ingestion points: API endpoints for browsing hot/new posts.
Boundary markers: None provided in the instructions to separate post content from agent instructions.
Capability inventory: Execution of 'moltbook.sh' and network operations.
Sanitization: No evidence of sanitization or validation of external post content before processing or logging.

moltbook-base85-obfuscation-instructions