The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly accesses the .env file to check for an 'Embedding API key'. Accessing .env files is a high-risk activity as these files typically contain sensitive secrets, tokens, and credentials.
[EXTERNAL_DOWNLOADS]: The skill performs remote software installation from external sources. It fetches the Bun runtime from bun.sh and installs the zod package via npm if they are not present or require repair. These are documented as well-known services.
[PERSISTENCE]: The documentation provides instructions to add the smoke test invocation to user startup scripts and login hooks, such as ~/.bashrc and ~/.zshrc. This is a persistence mechanism that ensures the skill's code runs automatically during new shell sessions.
[DYNAMIC_EXECUTION]: The skill is designed to automatically discover and execute all shell scripts located in the ~/.gbrain/smoke-tests.d/ directory. This allows for the execution of arbitrary code defined in local files at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local filesystem and external tool outputs, creating an attack surface for indirect injection.
Ingestion points: The skill loads and executes external shell scripts from ~/.gbrain/smoke-tests.d/*.sh and reads output from gbrain engine status --json.
Boundary markers: No explicit delimiters or instructions are used to isolate or ignore potentially malicious content within the loaded scripts or tool outputs.
Capability inventory: The skill utilizes the exec tool to run system commands and the read tool to inspect file contents.
Sanitization: There is no evidence of sanitization or validation performed on the scripts discovered in the user's home directory before execution.
[COMMAND_EXECUTION]: The skill uses the exec tool to run various system commands and 'auto-fix' scripts, including gbrain db-repair --yes, which can modify the system state without further user confirmation.

smoke-test