The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads and executes scripts from onecli.sh/install and onecli.sh/cli/install using curl | sh, which allows arbitrary remote code to run on the system.
[EXTERNAL_DOWNLOADS]: The skill fetches installation scripts and binaries from an external, non-trusted domain (onecli.sh) during the setup process.
[DATA_EXFILTRATION]: The skill reads the .env file to extract sensitive API keys for services like Anthropic and OpenAI, passing these credentials as arguments to an external binary.
[CREDENTIALS_UNSAFE]: The instructions target specific environment variables (ANTHROPIC_API_KEY, OPENAI_API_KEY) for extraction and migration, increasing the risk of credential exposure.
[COMMAND_EXECUTION]: The skill executes various system commands to modify shell initialization files (~/.bashrc, ~/.zshrc) and interact with the Docker daemon.
[PROMPT_INJECTION]: The skill processes untrusted user input to register secrets, creating a potential injection surface.
Ingestion points: User chat responses for keys/tokens in Phase 3.
Boundary markers: Absent.
Capability inventory: onecli secrets create (subprocess execution).
Sanitization: Absent.

init-onecli