jolt: Skill for Claude Code

This skill handles all the boilerplate: creating the objective struct, implementing the Objective trait, registering it in the appropriate enum, creating a const key, adding an ObjectiveFunction, creating a Criterion benchmark, and running sync_targets.sh.

<Execution_Policy>

• The user must provide an objective name (lowercase with underscores, e.g. cyclomatic_complexity).
• Use the request, spec, and repository context to identify the measurement and objective kind; ask only for missing requirements.
• Follow existing patterns exactly — study lloc.rs (static analysis) and binding.rs (performance) as models.
• Always run clippy and tests before reporting success. </Execution_Policy>

Phase 1: Gather Requirements

1. Validate the argument {{ARGUMENTS}}: must be a valid Rust identifier (lowercase alphanumeric + underscores). Reject otherwise.
2. Gather these requirements from existing context; ask the user only for missing information:
   • What is being measured? (becomes the description())
   • Is this a static analysis objective or a performance objective?
     • Static analysis: computes a metric by analyzing source code (e.g. lines of code, complexity). Overrides collect_measurement(), uses Setup = ().
     • Performance: computes a metric by running/profiling some computation (e.g. wall-clock time, max RSS). Overrides setup() and run().
   • What are the units? (e.g. "lines", "s", "bugs")
   • What files/directories does this objective measure? (used for diff_paths() scoping)

Phase 2: Explore Context

1. Read jolt-eval/src/objective/mod.rs to understand the current enums and dispatch methods.
2. Read jolt-eval/src/objective/objective_fn/mod.rs to understand objective function registration.
3. Read an existing objective for reference:
   • Static analysis: jolt-eval/src/objective/code_quality/lloc.rs
   • Performance: jolt-eval/src/objective/performance/binding.rs
4. Explore the modular crates that own the measured functionality.