The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages GrowthBook API credentials and implements robust measures to prevent accidental data exposure. It enforces masking of API keys in all terminal and chat outputs, showing only the last four characters. Additionally, it provides a transcript-exposure warning to users before they paste sensitive keys into the chat interface.\n- [COMMAND_EXECUTION]: The skill utilizes the Bash tool for configuration management and credential validation. It follows a secure implementation order by locking down the configuration directory with owner-only access (chmod 700) before sensitive data is written, and subsequently ensuring the configuration file itself is restricted (chmod 600).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied configuration data and incorporates an evidence chain for indirect injection surface analysis:\n
Ingestion points: User input for GB_API_KEY and GB_API_URL (SKILL.md).\n
Boundary markers: The skill instructs the agent to perform strict shape validation on inputs rather than relying solely on delimiters.\n
Capability inventory: Shell access via the Bash tool for file management and API validation calls (SKILL.md).\n
Sanitization: The skill enforces validation checks for the GB_API_URL, requiring the HTTPS protocol and prohibiting path components to mitigate potential request smuggling or misrouting.\n- [SAFE]: The skill's behavior is entirely consistent with its stated purpose of configuring GrowthBook access. Network operations are directed only at the vendor's official domain or a user-specified endpoint for the legitimate purpose of authentication validation.

gb-setup