Skill UI

This tool serves as a persistent memory and state manager for complex coding projects within the Claude environment. It allows users to initialize new projects with structured metadata, track detailed session progress (including summaries, decisions, and next steps), and save the entire context reliably. By ensuring project continuity across disconnected sessions, it drastically improves the flow and consistency of the development workflow.

This technique audits user interaction flows (like button clicks) to detect subtle bugs that static analysis often misses. It specifically tracks the full sequence of state changes, identifying instances where later actions unintentionally cancel, contradict, or leave the UI in an inconsistent state due to state store side effects or race conditions. Ideal for debugging complex state management architectures (Redux, Zustand).

This tool enables the definition and management of structured cognitive patterns for advanced AI agents. It guides decision-making by codifying complex reasoning strategies, including decision trees, chain of thought (CoT), and analogy mapping. Use it to build robust, self-improving agents that can solve complex problems step-by-step and track their learning status.

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.

Pattern for controlling LLM API costs while preserving quality. Combines complexity-based model routing, immutable budget tracking, retry logic limited to transient failures, and prompt caching. Use when building LLM API apps, processing batches with varying difficulty, or enforcing budget guardrails in production.

This skill monitors production cost burn rate trends over time by binning session spend into configurable buckets. It calculates window-over-window deltas and triggers exit code 1 if the latest bucket exceeds the prior mean by a configurable percentage. Useful for CI integration to detect spending acceleration before budget limits are hit, distinct from benchmark drift tracking.

Lists every session in cost-tracking with start time, message count, top model, and total cost. Use it for per-conversation spend analysis, retrospectives, and project rollups.

This skill exports cost-tracking telemetry data in Prometheus textfile or webhook JSON formats for external observability systems like Grafana, Datadog, and custom dashboards. It pulls session and budget records from the cost-tracking namespace, enabling real-time metric updates and ad-hoc reporting. Supports cron scheduling for near-real-time dashboards and one-shot webhook calls for immediate notifications.

Computes USD-per-day burn rate from recent cost-tracking sessions and linearly projects spend to 7d/30d/90d/365d horizons, including days until budget exhaustion.

Reads the active Claude Code session jsonl, tallies per-model token usage, computes USD cost using reference pricing, and persists a structured record to the cost-tracking AgentDB namespace for reporting and optimization.

This skill analyzes and reports on the token usage, expenditure, and budget for AI models (like Claude Code) by querying a local SQLite database. It is essential for users who need detailed financial accountability, tracking costs by project, tool, or date, and managing operational budgets.

This skill analyzes benchmark run data stored in JSON files to track performance drift over time. It calculates deltas between the first and last runs for metrics like win rate, latency, escalation rate, and LLM-baseline cost. The tool helps identify regressions before release, after corpus expansion, or following agent upgrades by flagging significant changes in speedup or latency compared to baselines.