The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install the nib CLI by downloading a shell script from https://nibrun.com/install.sh and piping it directly into the shell (| sh). This is a high-risk pattern as it executes unverified code from a remote source on the host system.\n- [REMOTE_CODE_EXECUTION]: The skill provides instructions for using the nib run command with remote HTTPS URLs to deploy and execute binaries. This pattern allows for the execution of arbitrary remote code within the nibrun environment.\n- [COMMAND_EXECUTION]: The agent is prompted to execute numerous shell commands for building binaries (e.g., bun build --compile), logging into services, and managing deployments. This requires the agent to have significant shell access and execution privileges.\n- [CREDENTIALS_UNSAFE]: The documentation suggests passing sensitive environment variables, such as API keys (e.g., STRIPE_SECRET_KEY), as command-line arguments using the --env flag. This practice is insecure as it can expose secrets in the system's process list, shell history, and logs.\n- [EXTERNAL_DOWNLOADS]: The skill downloads binaries and data archives from external sources, including GitHub releases and user-provided URLs, without verifying the integrity of the downloaded content.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection because it retrieves and processes potentially untrusted data from external sources.\n
Ingestion points: Application logs accessed via the read_logs tool or nib apps logs command, and SQLite database content accessed via query_database or nib apps sqlite.\n
Boundary markers: No delimiters or defensive instructions are present to prevent the agent from interpreting text within logs or databases as commands.\n
Capability inventory: The skill possesses extensive capabilities, including command execution, network access to the nibrun API, and local file system access for building and uploading artifacts.\n
Sanitization: There is no evidence of sanitization or filtering of the data ingested from logs or databases before it is returned to the agent context.

deploy-to-nibrun