|
1
|
--- |
|
2
|
name: code_review |
|
3
|
description: Spawn a code review (architect) subagent for deep analysis, planning, and debugging. The architect specializes in strategic guidance rather than implementation. Architect should be called after building major features. Relies on `delegation` skill. |
|
4
|
--- |
|
5
|
|
|
6
|
# Architect Skill |
|
7
|
|
|
8
|
Spawn an code review (a.k.a architect) subagent for analysis and planning. The architect specializes in analysis and strategic guidance rather than implementation. |
|
9
|
|
|
10
|
## When to Use |
|
11
|
|
|
12
|
Use this skill when: |
|
13
|
|
|
14
|
- You need deep architectural analysis or code understanding |
|
15
|
- You want strategic recommendations about system design or patterns |
|
16
|
- You need comprehensive analysis of code quality or technical debt |
|
17
|
- You want root cause analysis and debugging assistance |
|
18
|
|
|
19
|
## When NOT to Use |
|
20
|
|
|
21
|
- Simple tasks that you can complete directly |
|
22
|
- Tasks that require file edits or implementation (use delegation skill instead) |
|
23
|
- Read-only operations (use ripgrep or glob/read tools instead) |
|
24
|
|
|
25
|
## Available Functions |
|
26
|
|
|
27
|
### architect(task, relevantFiles, ...) |
|
28
|
|
|
29
|
Spawn an architect subagent for analysis and planning. |
|
30
|
|
|
31
|
**Parameters:** |
|
32
|
|
|
33
|
- `task` (str, required): The analytical task or question for the architect |
|
34
|
- `relevantFiles` (list[str], required): Full file paths to analyze |
|
35
|
- `responsibility` (str, default "evaluate_task"): Focus area: "debug", "plan", or "evaluate_task" |
|
36
|
- `includeGitDiff` (bool, default False): Include current unstaged git diff |
|
37
|
- `relevantGitCommits` (str, optional): Git commit range to analyze (e.g., "HEAD~3..HEAD") |
|
38
|
|
|
39
|
**Returns:** Dict with analysis results |
|
40
|
|
|
41
|
```json |
|
42
|
{ |
|
43
|
"success": true, |
|
44
|
"message": "Analysis summary", |
|
45
|
"subagentAlias": "architect_1", |
|
46
|
"result": "Full analysis output..." |
|
47
|
} |
|
48
|
``` |
|
49
|
|
|
50
|
**Responsibilities:** |
|
51
|
|
|
52
|
- **evaluate_task**: Assess completed or ongoing work against goals |
|
53
|
- **plan**: Create implementation plans with task decomposition and sequencing |
|
54
|
- **debug**: Root cause analysis, reproduction steps, and recommended fixes |
|
55
|
|
|
56
|
**Example:** |
|
57
|
|
|
58
|
```javascript |
|
59
|
// Plan a new feature |
|
60
|
const result = await architect({ |
|
61
|
task: "Create a plan for implementing rate limiting on API endpoints.", |
|
62
|
relevantFiles: ["src/middleware/index.ts", "src/routes/api.ts"], |
|
63
|
responsibility: "plan" |
|
64
|
}); |
|
65
|
console.log(result.result); |
|
66
|
|
|
67
|
// Debug an issue |
|
68
|
const result2 = await architect({ |
|
69
|
task: "The UserAuthService.validateSession() returns false for valid tokens.", |
|
70
|
relevantFiles: ["src/services/UserAuthService.ts", "src/utils/jwt.ts"], |
|
71
|
responsibility: "debug", |
|
72
|
includeGitDiff: true |
|
73
|
}); |
|
74
|
``` |
|
75
|
|
|
76
|
## Best Practices |
|
77
|
|
|
78
|
1. **Be specific in your task description**: Include concrete function names, error messages, or design goals |
|
79
|
2. **Provide relevant files**: The architect can only analyze files you pass in `relevantFiles` |
|
80
|
3. **Choose the right responsibility**: Use "plan" for new work, "debug" for issues, "evaluate_task" for reviewing progress |
|
81
|
4. **Use `includeGitDiff`**: When debugging regressions, include the diff to help the architect identify recent changes |
|
82
|
5. **Use `relevantGitCommits`**: When you need the architect to understand recent history (e.g., "HEAD~3..HEAD") |