③ Run your first code review

Goal

Use combo agent to review a real pull request and deliver:

• L1 static analysis (cppcheck / clang-tidy)
• L2 semantic review (MISRA-C:2012–oriented)
• Auto-post back to GitLab / Gerrit / GitHub

Prerequisites

• An active PR on GitLab / Gerrit / GitHub (C / C++)
• An access token for the target forge

Steps

Step 1 — Configure code hosting (~5 minutes)

Dashboard → Integrations → [your platform] → add token.

Step 2 — Create a review session (~1 minute)

Create a Combo named PR#123 review.

Step 3 — Kick off review (~15 minutes)

/code-review Review https://gitlab.example.com/.../merge_requests/123
Using MISRA-C:2012, output actionable fixes.

The system:

1. Fetches PR diff
2. Runs L1 static analysis (cppcheck / clang-tidy / semgrep)
3. Runs L2 LLM-assisted semantic review
4. Consolidates structured findings

Step 4 — Triage findings (~5 minutes)

Each item includes:

• Severity (Critical / Major / Minor)
• Rule ID (MISRA-C Rule X.X.X)
• Suggestion (code snippet)
• Confidence

Select items and bulk “post as comments”.

Step 5 — Post back to PR (~5 minutes)

/gitlab Post current review results to MR #123

Or GitHub / Gerrit:

/github Submit results as review comments on PR #123

Acceptance checklist

• Session captures both L1 and L2
• Comments visible on the PR UI
• Every item has severity and rule reference
• You decided on each suggestion (accept / ignore / revise)

Next steps

• Webhook-triggered PR review
• Automotive code-review solution

FAQ

Q: How are static analyzers configured? A: Bundled defaults—no manual install. On-prem, ops can toggle tools from the admin panel.

Q: Python / Java? A: Yes—rule sets adapt (PEP 8 / PMD / SpotBugs, etc.).