Skip to main content

hackagent.attacks._lib.prompt_parser

Helpers for parsing attacker outputs into prompt data.

extract_prompt_and_improvement​

def extract_prompt_and_improvement(
        content: str,
        *,
        allow_plaintext: bool = True) -> Optional[Dict[str, str]]

Extract a prompt (+ optional improvement) from attacker output.

Supports direct JSON, JSON code blocks, regex extraction, and a plain-text fallback when no JSON structure is present.

extract_prompt​

def extract_prompt(content: str,
                   *,
                   allow_plaintext: bool = True) -> Optional[str]

Extract just the prompt string from attacker output.