omnifix-gitlab

OmniFix

Automated review finding fixer — triage with parallel subagents, sequential fixing, verification, thread resolution.

Fetch unresolved discussion threads from a GitLab MR, triage each finding with parallel subagents in isolated worktrees, present results for user approval, apply approved fixes sequentially, verify with a fresh-eyes agent, commit, post replies on threads, and clean up.

Core principle: Parallel triage + user approval gate + sequential fix + verification = safe, high-quality automated fixes.

Announce at start: "I'm using OmniFix to fix review findings on MR !{id}."

Prerequisites

• glab CLI authenticated (glab auth status to verify)
• Git repository with remote pointing to GitLab
• Current working directory is in the git repo
• MR must have unresolved discussion threads

Input Parsing

Accept any of: MR number (136), prefixed (!136), or full GitLab URL. Extract MR ID. If URL provided, extract project path and MR IID.

The Process

Input: MR number with unresolved review findings
    |
    v
Phase 1: GATHER — fetch unresolved discussions + MR data
    |
    v
Phase 2: TRIAGE — N parallel subagents validate findings in read-only worktrees
    |
    v
Phase 3: APPROVE — present triage results, user explicitly approves
    |               (NO code changes until approval)
    v
Phase 4: FIX — single subagent applies approved fixes sequentially in writable worktree
    |
    v
Phase 5: VERIFY — fresh-eyes verification subagent reviews all changes
    |
    v
Phase 6: COMMIT + POST — commit, push (with permission), reply on threads, resolve
    |
    v
Phase 7: CLEANUP — remove all worktrees (ALWAYS runs)

---

Phase 1: Gather

Fetch ALL data before dispatching triage agents.

Step 1: Fetch discussions.

mcp__omniforge__fetch_mr_discussions(mr_id="{id}", repo_root="{cwd}")

Returns structured discussion threads with: discussion_id, resolvable, resolved, type (inline/general), file_path, line_number, body, author, replies.