What problem does it solve?
Routine code transforms like converting var to const or removing console statements normally trigger paid LLM calls with hundreds of milliseconds of latency. This Skill executes those structural edits locally through the agent-booster WASM engine at sub-millisecond speed and zero token cost.
Core Features & Use Cases
- Deterministic Code Transforms: Executes six booster intents (var-to-const, add-types, remove-console, add-error-handling, async-await, add-logging) via
AgentBooster.apply() with measured average latency around 1.2 ms.
- Confidence-Gated Writes: Fails closed when confidence falls below 0.5, refusing to write the file and escalating the edit to a Tier 2/3 LLM route instead.
- Outcome Persistence: Records latency, confidence, strategy, and applied status to memory under the cost-tracking namespace and feeds the routing learner via hooks.
- Use Case: In a CI pipeline, bulk-convert hundreds of files from var to const with deterministic, reproducible output and no API billing, while logging each result for cost analysis.
Quick Start
Ask the agent to apply the var-to-const transform to a specific source file using the cost-booster-edit skill and report the confidence score.
