| name | change-model |
| description | Change Model skill template. Transforms code changes into structured change reports (WHY/WHAT/HOW/VALIDATION four-layer architecture), with call-chain checking, archival storage, and automatic Git analysis. Triggered when generating change reports, DiffLogs, Release Notes, or archiving change records. |
| model_tier | L1 |
| skill_tier | functional |
| composes | [{"atomic":"example-dev"},{"atomic":"example-code-map"}] |
| composed_by | [{"planning":"delegation"},{"meta":"skill-builder-guide"}] |
| context_budget | {"l1_metadata":125,"l2_body":5000,"l3_references":12000} |
| version | 1.2.0 |
| status | active |
| review_by | "2026-07-28T00:00:00.000Z" |
| trust_level | internal |
| requires_network | false |
| requires_file_write | true |
| compatibility | requires git, python3 |
| allowed_tools | Bash Read Write Edit Grep Glob |
| evolution | {"usage_count":0,"last_corrections":[],"stale_markers":[]} |
Change Model โ Change Model Skill Template
Positioning: Functional tier skill โ provides a structured change-report methodology (WHY/WHAT/HOW/VALIDATION), guiding the full workflow from requirements analysis through archival storage.
Composition: This skill orchestrates code-map (L0 file location) and dev (L1 spec lookup). It is orchestrated by delegation (planning tier).
Deep References: See references/archive-design.md for the archival system design; references/archive-workflow.md for the archival workflow; references/git-analysis.md for automatic Git analysis.
Trigger Conditions
- Generate change report / DiffLog / Release Notes
- Analyze changes from Git history / generate report from commits
- Archive change records / change archive / historical change query
- User asks "how to write a change report" / "how to archive changes"
- Create project-specific change-report skills
- Need call-chain checking methodology
Related Skills
- Skill Builder Guide โ meta tier skill creation
- Dev Specs โ atomic tier tech stack info
- Code Map โ atomic tier file location [L0]
- Reference implementation scripts in
scripts/ directory
1. Skill Overview
1.1 What Is Change Model
Change Model transforms raw code diffs into a four-layer structured document:
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Layer 1: WHY โ Change Context & Requirements โ
โ โโ 1. Change Background โ
โ โโ 2. Requirements Analysis โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ Layer 2: WHAT โ Impact & Risk โ
โ โโ 3. Impact Analysis โ
โ โโ 4. Risk Assessment โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ Layer 3: HOW โ Design & Implementation โ
โ โโ 5. Design Approach โ
โ โโ 6. Implementation Mapping โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ Layer 4: VALIDATION โ Verification & Delivery โ
โ โโ 7. Call-Chain Check (pre-test) โ
โ โโ 8. Test Verification โ
โ โโ 9. Delivery & Rollback โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
1.2 Problems Solved
| Problem | Change Model Solution |
|---|
| Diff fragmentation | Organize by causal chain, not by file listing |
| Lack of behavioral abstraction | Make "condition โ result" decision logic explicit |
| Hard for newcomers to understand | Four-layer structure: background โ requirements โ design โ verification |
| AI cannot infer intent | Structured, machine-readable section design |
| Changes cannot be traced | Archival system: date-based archiving + INDEX.md index |
| Missing historical reports | Git analysis: auto-generate report skeleton from commit history |
1.3 Distinction from Changelog Skills
| Dimension | Change Model (change-model) | Changelog (diffs) |
|---|
| Perspective | Forward-looking: designed before changes occur | Retrospective: recorded after changes occur |
| Granularity | One report per requirement/task | One entry per commit/PR |
| Content | WHYโWHATโHOWโVALIDATION complete causal chain | What changed, why it changed |
| Timing | Complete design before coding, complete verification before testing | Auto/manual recording after commit |
| Purpose | Guide development, risk control | History tracking, retrospective audit |
2. Layer 1: WHY โ Change Context & Requirements
2.1 Change Background
| Element | Description |
|------|------|
| **Requirement Source** | {requester/ticket number} |
| **Trigger Reason** | {why this change is needed} |
| **Desired Goal** | {what effect the change should achieve} |
Fill-in Guidelines: Source must be specific and traceable / Reason must clarify "what the problem is" / Goal must be verifiable.
2.2 Requirements Analysis
Input: {input params}
Judgment conditions:
- {condition 1} โ {result 1}
- {condition 2} โ {result 2}
Output: {output params}
Fill-in Guidelines: Use "condition โ result" format / Cover all business branches / Specify inputs and outputs explicitly.
3. Layer 2: WHAT โ Impact & Risk
3.1 Impact Analysis
| Dimension | Impact Description | Impact Level |
|------|----------|:--------:|
| **Upstream Callers** | {description} | None/Low/Med/High |
| **Downstream Dependencies** | {description} | None/Low/Med/High |
| **Database** | {tables involved} | โ |
3.2 Risk Assessment
| Risk Item | Level | Description | Mitigation |
|--------|:----:|------|----------|
| {risk name} | L0/L1/L2/L3 | {description} | {mitigation} |
Risk Level Definitions:
| Level | Definition | Example |
|---|
| L0 | No external impact | Internal refactor, log optimization |
| L1 | Internal API change | New optional parameter, internal interface adjustment |
| L2 | External contract change | Interface signature change, response format change |
| L3 | Data/state migration | Database migration, config change |
4. Layer 3: HOW โ Design & Implementation
4.1 Call Chain
{Caller}
โ {Request method} {endpoint path}
โผ
Interface Layer ({entry component})
โ {business layer method}
โผ
Business Layer ({business component})
โ {business logic description}
โผ
Data Layer ({data component})
โ {query/persistence description}
โผ
Database
4.2 Implementation Mapping
#### Changed File List
| # | Operation | File Path | Change Description |
|:----:|:----:|----------|----------|
| 1 | โญ/โ๏ธ | `{path}` | {description} |
5. Call-Chain Checking Methodology
5.1 Core Objectives
| Objective | Description |
|---|
| Chain completeness | No breaks from entry point to endpoint |
| Type matching | Data types correctly passed at each link |
| Final call | SQL/API/message etc. final operation is correct |
5.2 Checking Procedure
Step 1: Identify the entry point
โโ API endpoint / page event / scheduled task / message consumption
โโ Record input parameter types and validation rules
Step 2: Trace intermediate links
โโ Interface layer โ Business layer โ Data layer
โโ Record method signatures and return types at each link
โโ Check whether types match
Step 3: Verify the final call
โโ SQL statement / external API / message queue / cache write
โโ Check whether parameters are passed correctly
Step 4: Error handling check
โโ Are exceptions caught?
โโ Are errors returned correctly?
5.3 Checklist
| # | Check Item | What to Check | Verification Method |
|---|
| 1 | Entry parameters | Parameter types, required-field validation | Inspect interface definition |
| 2 | Parameter passing | Are parameters passed correctly across layers? | Trace the method call chain |
| 3 | Type matching | Do input/output types match at each layer? | Compare type definitions |
| 4 | Null handling | Do optional fields have default values? | Inspect code logic |
| 5 | Final call | Are SQL/API/message parameters correct? | Inspect final implementation |
| 6 | Return handling | Is the return value assembled correctly? | Inspect return path |
| 7 | Exception handling | Are exceptions caught? | Inspect try-catch |
| 8 | Degradation logic | Is there a fallback on failure? | Inspect fallback |
5.4 When to Check
| Phase | Execute? | Notes |
|---|
| After requirements analysis | โ | Not yet implemented |
| After coding | โ
| Must execute |
| Before testing | โ
| Ensure chain is correct before testing |
| Before launch | โ
| Final verification |
5.5 Call-Chain Check Template
### 7. Call-Chain Check
| Check Item | Status | Notes |
|--------|:----:|------|
| Entry validation | โ
/โ | Parameter validation correct |
| Type check | โ
/โ | Data types match |
| Final call | โ
/โ | SQL/API call correct |
**Data Flow Verification**:
{Entry}
โ {Intermediate link 1}
โ {Intermediate link 2}
โ {Final call}
Pass Condition: All check items must be โ
.
6. Generating a Project-Specific Skill
6.1 Customization Items
| Customization | Description |
|---|
| Section add/remove | Add or remove sections as needed for the project |
| Risk levels | Adjust L0-L3 definitions to fit the project |
| Template format | Markdown / JSON / YAML |
| Trigger words | Project-specific change-related vocabulary |
6.2 Merging User Documentation
If the user provides an existing change report template or spec document, the user's document takes precedence:
1. Read the change report template/spec provided by the user
2. Compare differences with this text template
3. Use the user's document as the baseline; supplement missing parts with this methodology
4. Record differences for user confirmation
| Difference Type | Handling |
|---|
| User template has unique sections | Keep user sections, mark as project customization |
| User template is missing a layer | Suggest adding it (e.g., missing VALIDATION) |
| User has different risk level definitions | Use user's definitions |
| User uses different terminology | Keep user's terminology |
7. Archival System
7.1 Main Flow
Coding complete โ Call-chain check โ Test verification โ Generate final report โ ๐ฆ Archive
7.2 Directory Structure
docs/changes/
โโโ INDEX.md # Master index (dual-dimension: by time / by type)
โโโ 2026-04-26/
โ โโโ 20260426-01-add-filter-feature.md
โ โโโ 20260426-02-fix-login-timeout.md
โโโ 2026-04-27/
โโโ 20260427-01-refactor-payment-module.md
7.3 Archive Validation Checklist
7.4 Reference Scripts
python skills/change-model/scripts/generate-git-report.py \
--range "HEAD~3..HEAD" --title "Add filter feature" --type feature --archive
python skills/change-model/scripts/archive-report.py \
path/to/report.md --title "Add filter feature" --type feature --scope api
See references/archive-design.md for archive system design.
See references/archive-workflow.md for archive workflow.
See references/git-analysis.md for Git automatic analysis.
8. Model Tier
L1 โ Sonnet / functional tier: Requires understanding change semantics, organizing causal chains, and generating structured documents. File lookups are delegated to L0 โ Haiku.