forge-guardrails 0.1.2

Foundation types for an LLM-agent workflow framework
Documentation
# Prompts Development Guidelines

When modifying files under [src/prompts](file:///Users/whit3rabbit/Documents/GitHub/forge-rs/src/prompts), keep templates and parsers stable by adhering to the following rules:

## Gotchas & Rules

1. **Strict Prompt Parity**:
   - The text blocks generated by [nudges.rs]file:///Users/whit3rabbit/Documents/GitHub/forge-rs/src/prompts/nudges.rs are compared against Python test fixtures byte-for-byte in parity tests.
   - Even minor changes to spacing, capitalization, or wording will break `cargo test --test parity_tests`.
   - If prompt updates are required, you **MUST** update the Python reference generator first, recreate the JSON fixtures, and then match the exact output in Rust.
2. **Robust Multi-Strategy Parser**:
   - [parse_strategies.rs]file:///Users/whit3rabbit/Documents/GitHub/forge-rs/src/prompts/parse_strategies.rs must tolerate models wrapping tool arguments in markdown codeblocks (e.g., ` ```json ... ``` `), extra leading/trailing whitespace, and malformed trailing braces.
   - Do not simplify parsing logic in a way that breaks recovery of dirty text blocks.

## Testing Target

When modifying prompts or parsers, run:

```bash
cargo test --package forge-guardrails --lib prompts
cargo test --test parity_tests
```