Plan to Hit GitLab Level 4

Step-by-Step Action Plan to Hit GitLab Level 4 (7,500 Points) Current Score: 3,868 points (Level 3) Target Score: 7,500 points (Level 4) Points Needed: 3,632 points Step 1: Cash In Your 10 Already-Open MRs (+1,100 Points) You already did 95% of the work on 10 open MRs—and all 10

have green pipelines (success). Don't leave these points on the table. gitlab!261178 (+110 pts) What to do: Delete the 1 noisy spec block at spec/lib/security/ci_configuration/secret_detection_build_action_spec.rb:189 that @ashvins asked to remove. Finish: Push commit, resolve thread, comment @gitlab-bot ready @ashvins.

gitlab!262103 (+110 pts) What to do: Apply @erran's exact inline suggestion on shellescape_ref (remove the 2 top-level regex constants, add the NUL check, inline [^A-Za-z0-9_\-.,:+\/@\n[:^ascii:]]) and add the sh -c "printf %s ..." round-trip test. Finish: Push commit, resolve @erran's threads, reply to @cbonk & @erran. gitlab!262558 (+110 pts) What to do: Add Changelog: added to the commit message, add 1 test checking the file is deleted from storage in cleanup_stale_spec_files_worker_spec.rb, paste @jnutt's SQL query plan into the MR description, and set work_item: https://gitlab.com/gitlab-org/gitlab/-/work_items/633765 in the YAML file. Finish: Push commit, resolve threads, ping @shameed3 & @jnutt.

gitlab!261030 (+110 pts) — Already approved by @ngala. Comment: @gitlab-bot ready or politely ping @mbenayoun for maintainer merge. gitlab!261025 (+110 pts) — Already approved by @jannik_lehmann. Comment @gitlab-bot ready to get a backend maintainer assigned to click Merge.

(Your other 5 open MRs — !4407, !4406, !262418, !261174, CNG!3156 — are already with reviewers/maintainers and will bring in another *~550 pts** as they merge.)* Score After Step 1: 3,868 + 1,100 = ~4,968 points (~2,532 points left) community-bonus Issues (+1,000 Points) Instead of doing ten small MRs, do 3 bonus-labeled issues where a single merge pays 230–430 points. Click to Self-Assign First (Rule: Never write code before assigning): gitlab#635913 — MCP tool to list pipeline schedules (community-bonus::200 = ~330 pts) gitlab#628563 — Fix concurrent mobile push subscription limit (community-bonus::100 = ~230 pts) gitlab#457538 — Show parent-child relationships in iteration report (community-bonus::300 = ~430 pts) Verify 0 Competing MRs: Already verified — none of these 3 have an open MR. Link the Issue in Every MR Description: Always write Closes #<issue_iid> in the MR description so you get the +30 pt linked-issue bonus on top of the bonus label. Score After Step 2: 4,968 + 990 = ~5,958 points (~1,542 points left) gitlab-org/gitlab (+1,040 Points) The GitLab MCP Server (app/services/mcp/ and lib/api/mcp/) is small, self-contained Ruby code with fast RSpec tests and the exact same AI Foundations reviewers every time.

Pick up these 4 unassigned MCP issues first (each worth ~130 pts): Assign #632547 — Accept query / search_term aliases in MCP search tool (15-line change + spec). Assign #635931 — List unknown and accepted parameter names in MCP validation errors. Assign #632622 — Allow MCP save_note to edit an existing comment when note_id is passed. Assign #604238 — Add create_merge_request_pipeline MCP tool.

Then do 4 more from the same MCP list (#635916, #635915, #635910, #635909). 8 Merged MCP MRs × 130 pts = +1,040 points. Score After Step 3: 5,958 + 1,040 = ~6,998 points (~502 points left) ai-assist, gitlab-lsp & Technical Posts Finish the last ~500 points using whichever of these two paths is fastest that week: Option A (4 Fast Code MRs in ai-assist / gitlab-lsp = +520 pts): Assign ai-assist#3375 — Module boundary migration ai/platform/flow_engine/ Assign ai-assist#3422 — Fix TypeError: string indices must be integers in DWS Assign ai-assist#3370 — Fix unrecognized message part ValueError in DWS Assign gitlab-lsp#3381 — Forward needed environment to sandboxed commands Option B (2 Technical Blog Posts + 1 MR = +530 pts): Submit 2 deep-dive GitLab technical articles (MCP Server / Duo CLI architecture) for +200 pts each (+400 pts total) + 1 merged MR (+130 pts). Score After Step 4: 7,518+ points → LEVEL 4 UNLOCKED 🏆 Every time you sit down for a session, run this exact 4-part loop: Morning Check (10 mins): Ask the agent: "Check my open GitLab MRs for any new reviewer comments or pipeline failures and fix them immediately." (Reviewers merge fastest when you reply within a few hours.) Claim 1–2 Issues: Click the contributors.gitlab.com/manage-issue link to self-assign before coding.

Local Test Before Push: Ruby (gitlab): bundle exec rubocop <files> and bundle exec rspec <spec> TypeScript (gitlab-lsp): npm test -- <test_file> Python (ai-assist): poetry run pytest <test_file> and poetry run flake8 Open MR & Request Review: Include Closes #<iid> in the description, wait for CI to turn green, then comment @gitlab-bot ready.

Plan to Hit GitLab Level 4

Plan to Hit GitLab Level 4

Key Takeaways

Related Articles

Deep Dive into OpenClaw AI: Architecture, Security, and the Shift to Deterministic Agents

Before you start: the whole picture of checking, and how to test without breaking production

DeepSeek Just Made AI Inference More Interesting 🚀

Prove Your Cyber Insurance Answers: Commands for the Questionnaire Controls

Discussion

Plan to Hit GitLab Level 4

Key Takeaways

Related Articles

Deep Dive into OpenClaw AI: Architecture, Security, and the Shift to Deterministic Agents

Before you start: the whole picture of checking, and how to test without breaking production

DeepSeek Just Made AI Inference More Interesting 🚀

Prove Your Cyber Insurance Answers: Commands for the Questionnaire Controls

Discussion

Related Articles

Dev.to
Deep Dive into OpenClaw AI: Architecture, Security, and the Shift to Deterministic Agents
The landscape of AI agents is shifting. For the past few years, the industry has been dominated by "cognitive loops"—autonomous agents that prompt themselves in a continuous cycle until a task is complete. While powerful, this approach often leads to the dreaded "infinite loop" or hallucinations whe

Dev.to
Before you start: the whole picture of checking, and how to test without breaking production
This is part 3 of the series "Protect Your Own Service Yourself — Self-Defense Security in the Age of AI." Do the checks you can yourself; when it's hard, EarthLink Network can run the assessment for you, and fix the gaps you can't fix on your own. Series contents Anyone can attack now, with AI — ch

Dev.to
DeepSeek Just Made AI Inference More Interesting 🚀
Everyone talks about bigger AI models. But DeepSeek is focusing on something just as important: how efficiently we run them. DeepSeek recently upgraded V4.1-Flash to V4.1-Flash-1017 and open-sourced DeepInfer, its high-performance inference framework. ⚡ Faster inference DeepInfer includes optimized

Dev.to
Prove Your Cyber Insurance Answers: Commands for the Questionnaire Controls
I swapped 2 of the usual tags (webdev, ai) for devops and startup because this piece is about infrastructure evidence. Swap back if you prefer. Someone will eventually ask you to confirm the “yes” boxes on a cyber insurance application. MFA everywhere. EDR on every host. Clean email authentication.