explain-changes-mental-model — Claude Code command: what it does
explain-changes-mental-model splits a diff into logical groups and gives an order to read them in, so a big commit or branch turns into something you can follow.
MIT ★ 2,707 updated 2026-03-05
.claude/commands/explain-changes-mental-model.mdWhat explain-changes-mental-model does
Point it at a commit, a branch, a range or custom, and it runs the matching git command itself: git diff HEAD for uncommitted work, git show for a hash, a three-dot diff against main or master for a branch. Then it sorts the touched files into groups by purpose, layer or feature area and works out which group leans on which.
The answer comes back as a short summary, the concepts worth knowing before you open the code, then the groups in review order with their files, the reason each change exists, and what it depends on. A closing section lists the parts that look odd but are deliberate. Formatting-only edits get dropped.
With no argument it only sees what is not committed yet, so a clean working tree gives you nothing to review. On a branch whose default is named something other than main or master, pass the range yourself. Diffs past 5000 lines are inspected with --stat first, which keeps the first pass at a high level.
What it helps with
- review someone else's commit without reading it top to bottom
- find where to start reading a long branch
- map dependencies between parts of a diff
- explain a sprawling pull request to the rest of the team
- spot the non-obvious parts of a big changeset
Example prompts for the agent
/explain-changes-mental-model abc1234— break this commit into groups and tell me which order to read them in./explain-changes-mental-model feature/my-branch— what changed against main, and where do I start?/explain-changes-mental-model custom— I'll paste the diff next, walk me through it.
How to install explain-changes-mental-model
Manually: put the files in place
.claude/commands/explain-changes-mental-model.mdCall it with /explain-changes-mental-model. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/explain-changes-mental-model.mdCall it with /prompts:explain-changes-mental-model. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/explain-changes-mental-model.prompt.mdManually: put the files in place
.gemini/commands/explain-changes-mental-model.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about explain-changes-mental-model
How do I install explain-changes-mental-model in Claude Code?
Copy the command file to .claude/commands/explain-changes-mental-model.md. It then runs as /explain-changes-mental-model.
What does explain-changes-mental-model do?
It groups the changes in a commit, branch or range by purpose, works out the dependencies between the groups, and hands you a reading order with the tricky parts flagged.
What can I pass as an argument?
A commit hash, a branch name, a range such as main..HEAD or origin/main...HEAD, or the word custom. Leave it empty and it looks at uncommitted changes.
Source text
explain-changes-mental-model.md · MIT
--- description: Build a mental model of changes by splitting them into smaller logical chunks argument-hint: [commit|branch|range|custom] --- # Mental Model Builder Analyze code changes and present them as a coherent mental model - breaking large diffs into logical chunks ordered by dependency so they can be understood incrementally. ## What This Command Does 1. **Gather changes** from the specified source (commit, branch, range, or custom input) 2. **Analyze and categorize** changes into logical groups by theme/purpose 3. **Identify dependencies** between change groups 4. **Present ordered explanation** so changes can be reviewed in a sensible sequence ## Usage ```bash # Current working changes (staged + unstaged) /mental-model # Specific commit /mental-model abc1234 /mental-model HEAD~3 # Branch comparison (vs main) /mental-model feature/my-branch # Explicit range /mental-model main..HEAD /mental-model origin/main...HEAD # Custom - user will provide diff/changes in next message /mental-model custom ``` ## Implementation Steps When this command is invoked: ### 1. Determine Change Source Parse the argument to determine what changes to analyze: **No argument provided:** - Run `git diff HEAD` to get all uncommitted changes (staged + unstaged) - If no changes found, report "No uncommitted changes to analyze" **Argument looks like a commit hash** (7-40 hex chars): - Run `git show <hash> --stat` first to verify it exists - Run `git show <hash>` to get the full diff **Argument looks like a branch name** (contains letters, no `..`): - Detect main branch: check if `main` or `master` exists - Run `git diff <main-branch>...<argument>` to compare **Argument contains `..`** (explicit range): - Use as-is: `git diff <argument>` **Argument is "custom":** - Tell the user: "Please paste or describe the changes you want me to analyze in your next message." - Wait for user input before proceeding ### 2. Extract the Diff Run the appropriate git command and capture the output: ```bash # Example for commit git show <hash> --no-color # Example for range git diff <range> --no-color ``` If the diff is very large (>5000 lines), use `--stat` first to get an overview, then selectively read the most important files. ### 3. Analyze and Categorize Changes Group changes into logical categories based on: - **Purpose**: What problem does this change solve? - **Layer**: API, UI, business logic, data, infrastructure, tests, config - **Feature area**: Authentication, payments, search, etc. - **Type**: New feature, refactor, bug fix, cleanup For each file changed, determine: - What logical group it belongs to - What other changes it depends on - What changes depend on it ### 4. Build Dependency Graph Identify how changes relate to each other: - **Foundation changes**: Types, interfaces, schemas that other code depends on - **Implementation changes**: Code that uses the foundation - **Integration changes**: Code that ties implementations together - **Test/config changes**: Usually depend on everything else Order groups so that: 1. Dependencies come before dependents 2. Lower-level abstractions come before higher-level ones 3. Core changes come before peripheral changes ### 5. Present the Mental Model Output the analysis in this format: ```markdown ## Mental Model: [Brief title of what these changes accomplish] ### High-Level Summary [2-3 sentences explaining the overall goal and approach of these changes] ### Key Concepts to Understand Before diving into the code, understand these concepts: - **[Concept 1]**: [Brief explanation] - **[Concept 2]**: [Brief explanation] - **[Concept 3]**: [Brief explanation] ### Change Groups (in review order) #### 1. [Group Name] - [Purpose] **Files:** `file1.ts`, `file2.ts` **Why this matters:** [Explain the purpose and context] **Key changes:** - [Change 1 and why] - [Change 2 and why] **Depends on:** Nothing (foundation) **Enables:** [What other groups need this] #### 2. [Group Name] - [Purpose] **Files:** `file3.ts`, `file4.ts` **Why this matters:** [Explain the purpose and context] **Key changes:** - [Change 1 and why] - [Change 2 and why] **Depends on:** Group 1 **Enables:** [What other groups need this] [Continue for all groups...] ### Review Sequence For the best understanding, review in this order: 1. **[Group 1]** - Start here because [reason] 2. **[Group 2]** - Then this because [reason] 3. **[Group 3]** - Finally this because [reason] ### Gotchas and Non-Obvious Details - [Any tricky parts or things that might be confusing] - [Implicit assumptions in the code] - [Things that look wrong but are intentional] ``` ## Important Notes - **Focus on understanding, not just describing**: Explain WHY changes were made, not just WHAT changed - **Use the user's domain language**: If the codebase has specific terminology, use it - **Be honest about complexity**: If something is genuinely complex, say so and break it down further - **Highlight the "aha moments"**: What insight makes everything click into place? - **Skip trivial changes**: Don't waste mental bandwidth on import reordering, formatting, etc. ## Error Handling If git command fails: - Report the specific error - Suggest alternatives (e.g., "Did you mean branch X?") If diff is empty: - Report "No changes found for the specified source" If changes are too large to analyze meaningfully: - Provide a high-level overview first - Offer to deep-dive into specific areas: "This is a large changeset. Want me to focus on a specific area? (e.g., 'the API changes' or 'the database migrations')" ## Example Output For a commit that adds user authentication: ```markdown ## Mental Model: Adding JWT-based User Authentication ### High-Level Summary These changes add user authentication using JWT tokens. The flow is: user submits credentials → server validates → server returns JWT → client stores and sends JWT with requests → server validates JWT on protected routes. ### Key Concepts to Understand - **JWT (JSON Web Token)**: A signed token containing user identity, passed in Authorization header - **Middleware pattern**: Express middleware that runs before route handlers to check auth - **Token refresh**: Mechanism to get new tokens without re-entering credentials ### Change Groups (in review order) #### 1. Auth Types & Interfaces - Foundation **Files:** `src/types/auth.ts` **Why this matters:** Defines the shape of auth data used everywhere else **Key changes:** - Added `User`, `TokenPayload`, `AuthResponse` types - These types ensure consistency across the auth system **Depends on:** Nothing **Enables:** All other auth code #### 2. JWT Utilities - Token Operations **Files:** `src/utils/jwt.ts` **Why this matters:** Core logic for creating and validating tokens **Key changes:** - `signToken()` - creates JWT with user data and expiration - `verifyToken()` - validates JWT and extracts payload **Depends on:** Auth types **Enables:** Auth middleware, login endpoint [etc...] ```
© softaworks · MIT · softaworks/agent-toolkit