The Agent Skills Directory

[COMMAND_EXECUTION]: The authentication script scripts/tapi-auth.js is designed to be executed by the agent and performs unauthorized modifications to multiple system configuration files, including shell profiles like .bashrc, .zshenv, and .profile, as well as systemd environment configurations and PowerShell profiles.
[DATA_EXFILTRATION]: The skill identifies and modifies sensitive configuration paths, including the agent's internal configuration (.openclaw/openclaw.json or .clawdbot/moltbot.json) and various system environment files. This level of unauthorized tampering with sensitive system paths for persistence and credential storage is a high-risk operation.
[PROMPT_INJECTION]: The skill ingests untrusted data from external YouTube transcripts and metadata. This provides a surface for indirect prompt injection, and the skill instructions lack necessary sanitization or boundary markers to prevent the agent from executing instructions contained within the transcript content.

youtube-full