dmzoneill Review PR Multi-Agent Test
Test skill for multi-agent review. Runs fewer agents, doesn't post by default.
Inputs
| Input | Type | Default | Purpose |
|---|---|---|---|
mr_id |
int | required | GitLab MR ID |
agents |
string | "architecture,security,performance" | Comma-separated |
post_combined |
bool | false | Post to MR (default: no) |
model |
string | "sonnet" | Model |
Workflow
1. Load Persona
persona_load("developer")
2. Get MR Diff
gitlab_mr_diff(project="automation-analytics/automation-analytics-backend", mr_id)
3. Test Architecture Agent
- Run
claude --model sonnetwith prompt: "Review this code for architecture issues: {diff[:500]}" - Capture stdout/stderr, returncode
4. Test Security Agent
- Run
gemini --model sonnetwith prompt: "Review this code for security issues: {diff[:500]}" - Capture stdout/stderr, returncode
5. Build Summary
- Architecture: error, review preview
- Security: error, review preview
6. Log
memory_session_log("Skill completed", ...)
Output
Summary with agent results (errors, review snippets).