Skill Claude CodeCodex
Critically compare a feature architecture with a simpler option and a more scalable option, checking responsibility boundaries, dependency direction, testability, operational cost, and premature abstraction. Use for architecture reviews, technology choices, ADRs, or “is there another way?” questions.
Skill Claude CodeCodex
Coach debugging through symptom isolation, user-authored hypotheses, minimal verification experiments, and evidence-backed narrowing, then record the attempt in Learning MCP. Use when a feature has an error, failing test, unexpected behavior, or the user wants to train debugging ability.
Skill Claude CodeCodex
Review a Learning MCP feature or staged commit using compact evidence, separating human and AI ownership, tracing code flow, challenging architecture choices, and recommending up to three study topics. Use for feature completion, commit review, learning review, or AI-dependence assessment.
Skill Claude CodeCodex
Start and finish feature-based developer learning sessions with explicit goals, success conditions, human-owned decisions, and AI-allowed work. Use when the user starts a feature, asks to track a coding task, or wants the Learning MCP to measure a development session.
Skill Claude CodeCodex
Turn verified Learning MCP feature reviews into evidence bundles for the existing book writer and read-only reviewer. Use when the user wants to convert development experience, architecture decisions, or debugging cases into an Obsidian book chapter candidate.
MCP server Claude CodeCodexCursor +2
MCP server "learning-mcp" as configured in giseungNoh/AI-learning-mcp. Runs locally from the learning-mcp Python package.