A layered project memory system for coding agents, designed to preserve conventions, decisions, mappings, and session context for more consistent AI-assisted work.
Skill Claude CodeCodex
Generate structured git commit messages for staged changes using Conventional Commits plus Why/What/Risk/Tests sections. Use when preparing a commit after code changes, especially in repositories that keep living docs under memory-docs/. Prefer linking detailed rationale and progress to…
Skill Claude CodeCodex
Orchestrate a complete commit flow: use commit-planner to define batches, stage the selected batch, generate the message with commit-messages, and execute git commit safely.
Skill Claude CodeCodex
Analyze staged, unstaged, and untracked changes; group them into functionally coherent commit batches; and recommend staging order, commit boundaries, and related memory-docs references. Use when the working tree contains multiple kinds of changes and Codex needs to decide how many commits to make, which files belong…
Skill Claude CodeCodex
Use at the start of a conversation to establish repository context, or when the user asks to review code changes, compare branches, or trace file and line history. Covers three directions: git-scan (repo overview), git-review (understand a concrete change), and git-trace (follow historical evolution).
Skill Claude CodeCodex
Initialize or migrate an installed memory-docs system from repository evidence. Use after memory-docs templates are installed but still contain placeholders, when existing briefs, plans, notes, and docs must be distilled into the standard layers, or when a user explicitly asks to initialize installed project memory.…
Skill Claude CodeCodex
A workflow guide for research-focused Python projects. It covers data processing, regression analysis, diagnostics, tests, result tables, and matplotlib charts.
Skill Claude CodeCodex
Maintain an initialized memory-docs system after meaningful implementation, research, operational, documentation, decision, milestone, or handoff work. Use when verified durable facts or continuation context should be distilled into canonical owners, or when a user asks to audit, compress, archive, or repair project…