Skill UI

An advanced system that analyzes coding sessions via hooks to extract atomic, confidence-scored behaviors called "instincts." It tracks patterns, allowing users to evolve these instincts into comprehensive skills, commands, or agents. V2.1 introduces project-scoped learning, preventing knowledge contamination between different codebases and improving development workflow efficiency.

This tool performs robust outlier detection on individual session spending. Utilizing Median and Median Absolute Deviation (MAD), it calculates a modified z-score, making it highly resistant to extreme values, unlike mean+sigma methods. It specifically identifies which singular sessions are anomalous, making it crucial for cost monitoring, auditing, and integrating into CI/CD pipelines to prevent unexpected resource overruns.

A high-performance utility for executing simple, deterministic code transformations (e.g., removing console logs, converting var to const). This engine leverages a WASM core to perform structural refactoring sub-millisecond, without calling any expensive LLM APIs. Ideal for CI/CD pipelines, automated bulk refactoring, and cost-sensitive development workflows where determinism and speed are paramount.

This tool routes tasks through a multi-tier system, specifically partitioning tasks into Agent Booster eligible (Tier 1) and standard LLM calls (Tier 2/3). It calculates and reports the estimated cost, leveraging the structural $0 cost of Booster bypasses for simple transforms, helping users optimize spend before running batches of similar tasks.

This skill monitors operational costs against defined budgets by reading accumulated spending and computing utilization. It emits a detailed 4-tier alert ladder (INFO, WARNING, CRITICAL, HARD_STOP) based on predefined thresholds (50%, 75%, 90%, 100%). It is essential for cost governance, ensuring resources are managed within budget limits, and preventing system overspending.

Monitors the acceleration trend of production spending (burn rate) over defined time windows. It compares the latest spending bucket against the historical average, providing an early warning system. This is crucial for detecting sudden, unexpected cost spikes, such as a 'hot loop,' independent of overall budget limits.

This skill wraps the `getCompactContext` functionality to efficiently retrieve a compacted ReasoningBank context, specifically optimized for cost-analysis queries. It generates a detailed report summarizing critical metrics such as tokens saved, memories retrieved, and agentic-flow availability. It is essential for monitoring and optimizing resource usage in complex AI workflows, providing deep visibility into the efficiency and cost of the context retrieval mechanism.

Performs multi-baseline counterfactual cost analysis by comparing actual session spending against hypothetical costs generated by fixed, non-optimal models (e.g., always-haiku, always-sonnet). This tool quantifies the financial savings realized by the smart routing mechanism, answering the critical question: 'Is the routing earning its keep?' It is ideal for quarterly cost reviews, setting CI performance gates, and validating routing configurations.

This composite CI gate runs four critical cost monitoring checks—budget, burn, anomaly, and projection—in parallel. It surfaces a single, combined health status by taking the maximum exit code. This tool is an operational entry point for integrating comprehensive resource spending alerts directly into CI/CD pipelines, ensuring that resource usage remains within defined boundaries and predicts potential budget overruns.

Automatically analyzes complex AI operations, including token usage patterns, model tier suitability, cache hit rates, and agent redundancy. It identifies potential waste and provides actionable, prioritized recommendations with estimated dollar savings to significantly reduce operational costs and improve efficiency across LLM pipelines.

This tool provides forward-looking cost projection by calculating the average daily burn rate from a specified measurement window. It linearly extrapolates future spending across multiple time horizons (7d, 30d, 90d, 365d) and alerts users to the projected date of budget exhaustion. It is essential for proactive financial planning, SRE cost monitoring, and setting CI/CD build cost gates.

This tool provides a granular, per-message cost breakdown within a single conversation session. It serves as a deep-dive utility following an anomaly detection, allowing operators to pinpoint exactly which messages incurred the highest costs. It reveals critical cost components like output tokens, cache writes, and model escalations, helping users understand the true resource consumption beyond simple token counts.