Code Analyzer
Analyze staged code and surrounding context for duplications, dead code, and optimization opportunities using GPT-5.3-Codex.
Usage
/code-analyzer → Codex high (default)
/code-analyzer deep → Codex xhigh
Argument Parsing
Keywords (reserved):
deep → Use xhigh effort instead of high
Parsing Logic
args = split(input)
effort = "high"
for arg in args:
if arg == "deep":
effort = "xhigh"
Result Mapping
| Input |
effort |
| (none) |
high |
deep |
xhigh |
Analysis Scope
| Scope |
Description |
| Staged files |
Production + test code in git staged |
| Surrounding code |
Same package/directory, related modules |
Analysis Items
| Item |
Description |
| Duplications |
Similar code patterns, copy-paste code |
| Dead code |
Unused functions, imports, variables |
| Optimizations |
Performance improvements, inefficient patterns |
| Consistency |
Pattern alignment with existing codebase |
Implementation
Invoke the 17th-code-analyzer subagent:
Task(
description="Analyze code quality ({effort})",
subagent_type="17th-code-analyzer",
prompt="""
Analyze code quality for staged changes and surrounding code.
Configuration:
- Effort: {effort} (high or xhigh)
- Project root: {project_root}
""",
run_in_background=true
)
Auto-Trigger
When skill auto-triggers (no args), uses defaults:
Output
- Report:
.kkachi/{branch_name}/completed/code-analyzer-{epoch_timestamp}.md
- Summary: duplications, dead code, optimizations found
- Recommendations: prioritized action items
Examples
| Input |
effort |
| (none) |
high |
deep |
xhigh |
Notes
- Uses GPT-5.3-Codex for all analysis
- Analyzes staged changes + surrounding code
- Supports Go, Python, Dart, TypeScript, JavaScript
- Results verified before reporting
- Ensure
.kkachi/ is in .gitignore
