Installs into .claude/skills of the current project.
Are you the author of Impact Analysis?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/seokan-jeong-impact-analysis)
---
name: impact-analysis
description: Use when you need to analyze cascade impact of changing a component.
user-invocable: false
---
# Impact Analysis Skill
Analyze the cascade impact of modifying a component using the project ontology's dependency graph.
## Usage
/team-shinchan:impact-analysis <target> [options]
## Parameters
- `<target>` — Entity name, ID, or file path to analyze
- `--depth <N>` — Traversal depth (default: 2, max: 5)
- `--direction outgoing|incoming|both` — Direction of analysis (default: both)
## Process
### 1. Locate Target Entity
1. Check if `.shinchan-docs/ontology/ontology.json` exists
2. If not: "No ontology found. Run `/team-shinchan:ontology scan` first."
3. Search for target by:
- Exact entity ID match
- Entity name substring match
- File path match (for Component entities)
4. If multiple matches, list them and ask user to clarify
### 2. Cascade Analysis
Run `node ${CLAUDE_PLUGIN_ROOT}/src/ontology-engine.js related <entityId>` with the specified depth.
Analyze the result:
- **Direct dependencies** (depth 1): Components that directly depend on target
- **Indirect dependencies** (depth 2+): Cascade through transitive relations
- **Test coverage**: TESTED_BY relations for affected components
- **Pattern compliance**: FOLLOWS_PATTERN relations that may be affected
### 3. Risk Assessment
Calculate risk level based on:
- **HIGH**: 10+ affected entities OR target has 5+ incoming DEPENDS_ON
- **MEDIUM**: 5-9 affected entities OR target has 3-4 incoming DEPENDS_ON
- **LOW**: 1-4 affected entities OR target has 0-2 incoming DEPENDS_ON
### 4. Output Format
Display:
```
Impact Analysis: {target name}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Risk Level: {HIGH|MEDIUM|LOW}
Direct Dependencies ({N}):
- {component name} ({file_path}) — {relation type}
...
Indirect Dependencies ({N}):
- {component name} ({file_path}) — via {intermediate}
...
Test Coverage:
- {test suite name} ({file_path}) — covers {component}
...
Affected Modules:
- {module name} ({N} components affected)
...
Recommendations:
- {recommendation based on risk level}
```
## Graceful Degradation
If ontology doesn't exist or target not found, suggest running scan first.