The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill performs a global installation of the GBrain engine directly from the author's GitHub repository using the command bun install -g github:garrytan/gbrain#latest-stable.
[COMMAND_EXECUTION]: The skill utilizes several CLI commands, including gbrain init, gbrain engine status, gbrain doctor, and gbrain connect, to manage the installation, repair, and verification of the memory system.
[CREDENTIALS_UNSAFE]: The skill handles sensitive information such as database connection strings and private credentials files (e.g., --credentials-file /absolute/private/agent-example.json). It explicitly instructs the agent to redact credentials in output and avoid copying secrets into persistent configuration files.
[EXTERNAL_DOWNLOADS]: The skill references and fetches various guides, documentation, and installation scripts from the garrytan/gbrain repository on GitHub.
[INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it facilitates the ingestion of untrusted external data such as chat history exports and personal notes into the agent's context.
Ingestion points: Capability to import notes and chat exports into the agent's memory as described in the optional capabilities section of SKILL.md.
Boundary markers: Absent; there are no explicit delimiters or instructions provided to the agent to treat the imported content as untrusted data.
Capability inventory: The skill has capabilities including shell command execution (gbrain CLI), file system writes (writing memory instructions), and network communication.
Sanitization: Absent; the skill does not specify any validation, filtering, or escaping mechanisms for the imported content.

setup