Module prompt 

Prompt template models.

Provides data structures for user-defined prompt templates with variable substitution.

Code Block Detection Edge Cases

Variable extraction automatically skips {{variable}} patterns inside fenced code blocks to avoid capturing documentation examples. This section documents edge cases and behaviors.

Supported Code Block Syntaxes

Syntax	Supported	Notes
```language ... ```	✓	Standard fenced code block
``` ... ```	✓	Code block without language
~~~ ... ~~~	✓	Tilde fenced code block
Indented code (4 spaces)	✗	Only fenced blocks detected

Edge Cases

1. Unclosed Code Blocks

Input: triple-backtick rust, let x = "{{var}}";, no closing triple-backtick

Behavior: Unclosed blocks are not detected, so variables inside ARE extracted. This is intentional - malformed content shouldn’t silently exclude variables.

2. Nested Code Blocks (within markdown)

Input: Outer tilde block containing inner backtick block with {{inner_var}}

Behavior: Both tilde and backtick blocks are detected. Variables inside either syntax are excluded. Nested blocks are handled correctly.

3. Variables at Block Boundaries

Input: {{before}} immediately before triple-backtick, {{after}} immediately after

Behavior: Both {{before}} and {{after}} are extracted. Only content strictly between the opening and closing triple-backticks is excluded.

4. Inline Code (single backticks)

Input: Use {{var}} syntax for variables. (single backticks around var)

Behavior: Single backticks DO NOT exclude variables. Only triple-backtick fenced blocks are detected. {{var}} IS extracted.

5. Empty Code Blocks

Input: Empty triple-backtick block

Behavior: Empty blocks are detected but contain no variables to exclude.

Workarounds

If you need a {{variable}} pattern in your actual prompt output (not as a variable):

1. Escape it: Use \{\{literal\}\} (will be preserved literally)
2. Put it in a code block: Variables in fenced blocks are not substituted
3. Use a variable with literal value: Define open_brace/close_brace variables

Macros

lazy_regex 🔒

Creates a compile-time verified regex wrapped in LazyLock.

Structs

CodeBlockRegion

Represents a fenced code block region in content.

ExtractedVariable

Result of extracting a variable from prompt content.

PromptTemplate

A user-defined prompt template.

PromptVariable

A template variable definition.

ValidationIssue

A validation issue found in prompt content.

ValidationResult

Validation result for prompt content.

Enums

IssueSeverity

Severity level for validation issues.

Constants

MAX_VARIABLE_VALUE_LENGTH

Maximum length for variable values (64KB).

RESERVED_PREFIXES 🔒

Reserved variable name prefixes that cannot be used.

Statics

CODE_BLOCK_BACKTICK_PATTERN 🔒

Regex pattern for detecting fenced code blocks (triple backticks with optional language identifier). Matches: followed by optional language, then content, then

CODE_BLOCK_TILDE_PATTERN 🔒

Regex pattern for detecting tilde fenced code blocks. Matches: ~~~ followed by optional language, then content, then ~~~

VALIDATION_PATTERN 🔒

Regex pattern for detecting any content between {{ and }}.

VARIABLE_PATTERN 🔒

Regex pattern for extracting template variables: {{variable_name}}. Supports dots for context template auto-variables like {{memory.id}}.

Functions

default_required 🔒

Default value for required field (true).

detect_code_blocks

Detects fenced code blocks in content.

extract_variables

Extracts variables from prompt content.

extract_variables_with_exclusions 🔒

Extracts variables from prompt content, excluding those inside code blocks.

is_in_exclusion

Checks if a byte position falls within any exclusion region.

is_reserved_variable_name

Checks if a variable name uses a reserved prefix.

sanitize_variable_value

Sanitizes a variable value to prevent template injection attacks.

substitute_variables

Substitutes variables in prompt content.

validate_prompt_content

Validates prompt content for common issues.