More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    GitLab CLI: glab Workflows and Automation

    1

    Automate GitLab with glab safely: batch MR and issue operations, a CI failure diagnosis recipe, the API escape hatch, stacked MRs, and credential rules for unattended runs.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    GitLab CLI: glab Workflows and Automation

    GitLab CLI: glab Workflows and Automation

    Example session with this skill installed

    Our GitLab project has 300+ open issues, half untouched for a year, and main-branch CI fails a few times a week with nobody triaging it. Set up the automation.

    • Read your context and instructions
    • Compiled the gitlab cli

    AUTOMATION SETUP — produced by applying the skill.

    Credentials first: a project token scoped to api, delivered as GITLAB_TOKEN in the environment, rotation calendared. The token value appears in no script, no log, no commit. Dashboards planned later will get a separate read_api token.

    Stale-issue job, built read-first. The selection query (open, untouched 90+ days, JSON output filtered with jq) runs as a dry run and selects 118 issues. A human reads a 20-issue sample before any write: two are quiet-but-alive tracking issues, so the query gains a label exclusion. First real run, scheduled while the team is awake: 116 issues commented ("Closing as stale — reopen if still relevant") then closed, every ID written to the run log. Three are reopened within a week — the comment-before-close rule working as designed, not a failure.

    CI diagnosis: the recipe runs on failed pipelines on the default branch — latest pipeline, failed jobs, last 60 lines of each log — and files the output with a classification. First month: 9 flaky-test flags (one test quarantined), 2 expired-secret incidents caught the morning they happened, 1 real breakage routed straight to its owner.

    Release notes: the changelog generator drafts from merged MRs; a human curates 34 draft entries down to 28 user-facing ones in about 20 minutes. Readiness rubric for the whole setup: 17/20, the missing points a second reviewer for the approval job — scheduled before that one runs unattended.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The web UI is fine for one merge request. It is not fine for the 300th stale issue, the nightly CI failure nobody has time to open, or release notes assembled by hand from a quarter of merged work. This skill is the automation companion to terminal GitLab use: glab inside scripts, cron jobs, and agent workflows, where the output has to be trustworthy because nobody is watching it live. It starts with credential handling for unattended runs — tokens in environment variables at minimum scope, never echoed, never logged, rotated on a schedule — because the token is what makes automation possible and what makes it dangerous. Batch operations are built on three rules that never bend: select with JSON output and jq rather than scraping tables, dry-run every selection query and read the IDs before any write, and comment before closing so humans can object. A complete CI failure diagnosis recipe (latest pipeline, failed jobs, log tails) comes with a classification table that lets a scheduled job or agent sort flaky tests from expired secrets from real breakage without further commands. The glab api escape hatch covers everything the CLI does not wrap, with the pagination rule that prevents the classic "script only processed page one" bug. Stacked MRs and release automation round it out: changelogs generated from merged MRs, then curated by a human, because automation's job is completeness and judgment stays yours. You get ready-to-adapt scripts for stale labeling, green-MR approval, and stale-issue closing, a runbook template, an eight-item pre-run checklist, a 20-point automation readiness rubric, and a worked example where the dry run catches two issues that must not be closed — before the job ever writes a thing.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.

    • glab wraps the GitLab API, not other forges: GitHub and Bitbucket have their own CLIs and different object models.
    • Bulk writes have no undo on GitLab's side. The skill's dry-run, comment-first, and logging rules reduce that risk; they cannot remove it, which is why a human reads the selection first.
    • Self-managed instances vary by version; a command or flag can differ on older installs, and the API escape hatch examples assume a reasonably current GitLab.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Software
    Git

    Permissions

    Terminal / Shell
    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    For a dozen items, you should — the skill says one-off tasks belong in the UI. Automation earns its risk when the operation recurs or scales past what clicking handles: hundreds of stale issues, a diagnosis that must run on every failed pipeline, release notes every single release.

    The selection discipline: JSON output plus jq filters, a mandatory dry run whose IDs a human reads, expected-count sanity checks per run, and comment-before-close so affected humans can object. The worked example's dry run caught exactly this — two tracking issues the filter would have wrongly closed.

    Under the same rules as any unattended job: minimum-scope token in the environment, read operations free, write operations only through reviewed scripts with dry-run records and run logs. An agent improvising bulk writes outside a reviewed script is outside what this skill sanctions.

    New in DevOps & Cloud