lilith-zero: Skill for Claude Code

Rust Bulletproof: Advanced Verification Strategy

This skill goes beyond standard unit testing to ensure the mathematical correctness and memory safety of Lilith's critical Rust path (lilith-zero/src). It employs fuzzing, property-based testing, and interpreter-based verification.

When to use this skill

• Critical Logic Changes: Modifications to policy_validator.rs, pattern_matcher.rs, or security.rs.
• Parsing/Input Handling: Changes to how external data (YAML, JSON, Byte streams) is ingested.
• Unsafe Code: ANY addition or modification of unsafe blocks.
• Concurrency: Changes involving Tokio, Mutex, or RwLock.

How to use it

1. The "No-Panic" Policy (Property Testing)

Use proptest to mathematically assert that functions never panic and properties hold for all inputs, not just example cases.

Command:

# Run property tests specifically (usually marked with #[test])
cd lilith-zero; cargo test --package sentinel --test properties

Note: If tests\properties.rs does not exist, ask the user to create a property test for the new feature.

2. Undefined Behavior Check (Miri)

If the code uses unsafe or does complex pointer arithmetic, you MUST run Miri. Miri interprets the code to find memory leaks, data races, and misalignment.

Command:

# Install Miri if missing
rustup component add miri
# Run tests under Miri
cd lilith-zero; cargo miri test

3. Fuzzing Campaign

For parsers and security boundaries, run a short fuzzing campaign to find edge case crashes.

Command:

# Ensure cargo-fuzz is installed
cargo install cargo-fuzz
# Initialize if needed: cargo fuzz init
# Run the relevant target (e.g., policy_parser)
cd lilith-zero; cargo fuzz run policy_parser -- -max_total_time=300 # 5 minutes

4. Mutation Testing (The "Test the Tests" Rule)

Verify that your tests are actually checking the code by introducing random bugs (mutants) and ensuring tests fail.