Implements new optimization objectives for the jolt-eval framework in Rust.
npx skills add https://github.com/a16z/jolt --skill new-objective
Please help me install this Agent Skill. Skill: new-objective Source: https://github.com/a16z/jolt/tree/main/.claude/skills/new-objective Command: npx skills add https://github.com/a16z/jolt --skill new-objective
Adding a new measurable objective to the jolt-eval optimization framework requires touching many files: the objective struct, trait implementation, enum registration, const keys, objective functions, and Criterion benchmarks. This Skill automates that boilerplate so a new objective is wired in correctly and consistently.
Ask the assistant to implement a new jolt-eval objective named cyclomatic_complexity that statically measures code complexity in the prover crate.
High-intent search queries and answers about installing and using this skill.
Provide a lowercase snake_case objective name and specify whether it is static analysis or performance. The Skill creates the objective file, implements the Objective trait, registers it in the enums, adds an ObjectiveFunction, and validates with clippy and tests.
Static analysis objectives compute metrics by analyzing source code, such as lines of code, and override collect_measurement() with Setup = (). Performance objectives run or profile computations, such as wall-clock time, and override setup() and run() measured via Criterion.
Yes. Performance objectives need a bench file in jolt-eval/benches that invokes the bench_objective! macro, followed by running sync_targets.sh to add the [[bench]] entry to Cargo.toml. Without it the optimization harness cannot measure the objective.
The objective name must be a valid Rust identifier using lowercase alphanumeric characters and underscores, such as cyclomatic_complexity. Names failing this validation are rejected before any files are created.
The Skill runs cargo fmt, clippy with warnings denied, and nextest for the jolt-eval package. Static analysis objectives are additionally checked via the measure-objectives binary, and performance objectives via a Criterion test-mode benchmark run.
OutlineDrivenCreates step-by-step implementation plans from codebase analysis and requirements.
danielmiesslerRun iterative optimization loops for code and AI models.
maschadProfile Rust code with Criterion benchmarks, TSC-based latency measurements, and flamegraphs.
cursorBuilds rerunnable codemods, scripts, and generators instead of performing edits by hand.
