CLI Adapters

CLI adapters implement the adapter pattern to normalize ForgeFlow's interaction with different AI CLI tools.

Interface

class CLIAdapter(ABC):
    def is_input_prompt(self, output: str) -> bool:
        """Detect if AI CLI is waiting for input."""

    def is_input_prompt_with_text(self, output: str) -> bool:
        """Detect if input prompt has text already entered."""

    def is_task_processing(self, history: list[str]) -> bool:
        """Detect if task is being processed (uses output history for change detection)."""

    def is_ai_cli_exist(self, output: str) -> bool:
        """Detect if AI CLI is running."""

    def wants_ansi(self) -> bool:
        """Whether to capture ANSI escape sequences (default False)."""

Built-in Adapters

CLI Type	File	Notes
claude_code	core/cli_adapters/claude_code.py	Default
gemini	core/cli_adapters/gemini.py	Full implementation
codex	core/cli_adapters/codex.py	Full implementation

Rule Files

Each CLI type has corresponding rules in core/cli_types/:

• core/cli_types/gemini_rules.py
• core/cli_types/claude_code_rules.py
• core/cli_types/codex_rules.py

Rules are loaded via forgeflow.rules.loader.get_rules(), which combines CLI-type rules from rules/builtin/ with task rules and custom user rules.