AI Engineering

Context Engineering helps you manage context windows in AI coding sessions by mastering four operations: Write, Select, Compress, and Isolate. Use it to keep token budgets efficient, plan compaction strategies, and partition work across subagents and worktrees.

Context Optimizer helps manage context windows and token budgets in Claude Code sessions. Use it when sessions feel slow, context is degraded, or you are running out of budget.

Cost Tracker monitors Claude Code session costs and token usage, sets budget alerts, and suggests strategies to optimize spend. Use it mid-session to check costs or set spending limits.

Domain Modeling builds the project's shared language and bounded contexts before writing code, so names stay consistent and the agent stops paraphrasing domain concepts. It produces a CONTEXT.md glossary and decision records, used at the start of a project or feature.

Learn Rule captures a correction or lesson as a persistent learning rule with category, mistake, and correction. Use it after mistakes or when the user says 'remember this', 'don't forget', 'note this', or 'learn from this'.

LLM Council orchestrates multi-LLM deliberation in three phases: independent responses, cross-model anonymized ranking, and chairman synthesis. Use it for high-stakes reviews, plan critique, or contested learning-rule resolution.

MCP Audit analyzes connected MCP servers for token overhead, redundancy, and security. Use when sessions feel slow or before adding new MCPs.

Module Map produces a one-screen map of an unfamiliar area of the codebase, showing entry points, modules, data flow, callers, and hidden coupling. Use it to orient quickly in unfamiliar code.

Orchestrate coordinates commands, agents, and skills to build features through structured research, planning, implementation, and review phases with validation gates. Use it for complex features touching multiple files or requiring architecture decisions.

Permission Tuner analyzes permission denial patterns and generates optimized alwaysAllow and alwaysDeny rules to reduce permission prompt fatigue. Use when permission prompts are slowing you down or after sessions with many denials.

Plan Interrogate stress-tests a plan by walking its decision tree one question at a time, resolving every open decision before code is written. Use it when a design needs pressure-testing before implementation.

Pro Workflow provides orchestration patterns, hook events, agents, and self-correction loops that reduce correction cycles in AI coding. Use it to set up CLAUDE.md memory, parallel worktrees, pre-flight discipline, and wrap-up rituals across Claude Code, Cursor, and other AI agents.