mirror of
https://github.com/Shiva108/ai-llm-red-team-handbook.git
synced 2026-07-11 06:53:40 +02:00
shiva108
ccb930d126
- Fixed MD040 errors in Chapter 14 (added language identifiers to 170 code blocks) - Fixed MD025 error in Chapter 15 (changed H1 to H3/H4 in example section) - Added RELEASE_REVIEW_REPORT.md documenting final audit - Created fix_md040.py automation script for future use All critical and high-priority linting issues resolved. Handbook is now ready for v1.0 Gold Master release.
116 lines
3.8 KiB
Python
116 lines
3.8 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Fix MD040 linting errors by adding language identifiers to bare code blocks.
|
|
Specifically designed for Chapter 14: Prompt Injection
|
|
"""
|
|
|
|
import re
|
|
import sys
|
|
|
|
def detect_code_block_language(content):
|
|
"""
|
|
Heuristically determine the language of a code block based on its content.
|
|
"""
|
|
lines = content.strip().split('\n')
|
|
|
|
# Check for Python indicators
|
|
if any(keyword in content for keyword in ['def ', 'import ', 'class ', 'print(', 'return ', ' #', 'if __name__']):
|
|
return 'python'
|
|
|
|
# Check for bash/shell indicators
|
|
if any(keyword in content for keyword in ['#!/bin/', 'echo ', 'curl ', 'grep ', 'sed ', 'awk ', '$ ', 'export ']):
|
|
return 'bash'
|
|
|
|
# Check for JSON indicators
|
|
if content.strip().startswith('{') and content.strip().endswith('}'):
|
|
if '"' in content and ':' in content:
|
|
return 'json'
|
|
|
|
# Check for YAML indicators
|
|
if re.search(r'^\w+:\s*$', content, re.MULTILINE) and not any(c in content for c in ['{', '(', 'def ']):
|
|
return 'yaml'
|
|
|
|
# Check for markdown content
|
|
if '## ' in content or '### ' in content or '- [ ]' in content:
|
|
return 'markdown'
|
|
|
|
# Check for prompts/examples (conversation format)
|
|
if any(indicator in content.lower() for indicator in ['user:', 'assistant:', 'system:', 'prompt:', 'response:']):
|
|
return 'text'
|
|
|
|
# Check for bullet points or list-like content
|
|
if content.count('\n- ') > 2 or content.count('\n* ') > 2:
|
|
return 'text'
|
|
|
|
# Check for HTTP/API content
|
|
if any(keyword in content for keyword in ['HTTP/', 'GET ', 'POST ', 'Content-Type:', 'Authorization:']):
|
|
return 'http'
|
|
|
|
# Default to text for plain examples
|
|
return 'text'
|
|
|
|
def fix_code_blocks(filepath):
|
|
"""
|
|
Read a markdown file and add language identifiers to bare code blocks.
|
|
"""
|
|
with open(filepath, 'r', encoding='utf-8') as f:
|
|
content = f.read()
|
|
|
|
# Pattern to match bare code blocks (``` without language identifier)
|
|
# This matches ``` at start of line, optionally followed by newline
|
|
pattern = r'^```\s*$'
|
|
|
|
lines = content.split('\n')
|
|
in_code_block = False
|
|
code_block_start = -1
|
|
code_block_lines = []
|
|
fixed_count = 0
|
|
|
|
i = 0
|
|
while i < len(lines):
|
|
line = lines[i]
|
|
|
|
# Check if this is a code fence
|
|
if re.match(r'^```+', line):
|
|
# Check if it's a bare fence (no language specified)
|
|
match = re.match(r'^(```+)(\s*)$', line)
|
|
|
|
if match and not in_code_block:
|
|
# Start of a bare code block
|
|
in_code_block = True
|
|
code_block_start = i
|
|
code_block_lines = []
|
|
i += 1
|
|
continue
|
|
elif match and in_code_block:
|
|
# End of code block - analyze and fix
|
|
code_content = '\n'.join(code_block_lines)
|
|
language = detect_code_block_language(code_content)
|
|
|
|
# Replace the opening fence with language identifier
|
|
lines[code_block_start] = f'```{language}'
|
|
fixed_count += 1
|
|
|
|
in_code_block = False
|
|
code_block_start = -1
|
|
code_block_lines = []
|
|
elif in_code_block:
|
|
code_block_lines.append(line)
|
|
|
|
i += 1
|
|
|
|
# Write back
|
|
with open(filepath, 'w', encoding='utf-8') as f:
|
|
f.write('\n'.join(lines))
|
|
|
|
return fixed_count
|
|
|
|
if __name__ == '__main__':
|
|
if len(sys.argv) < 2:
|
|
print("Usage: python fix_md040.py <markdown_file>")
|
|
sys.exit(1)
|
|
|
|
filepath = sys.argv[1]
|
|
count = fix_code_blocks(filepath)
|
|
print(f"Fixed {count} bare code blocks in {filepath}")
|