One-off jobs are nice, but the real return on a vibe-code agent is work that runs on its own — every night, every sprint, every merge. These ten automations chain Penpot together with the rest of your stack: Git, email, Slack, S3, and your CI pipeline. Describe each one, review the test run, and then let the schedule carry it.
1. Sync design tokens from your code repo
Many teams keep design tokens in the repository — a tokens.json from Style Dictionary or a Tailwind config — and then hand-copy values into Penpot until the two drift apart. The agent can close that loop with a scheduled job: read the repo file, apply colors and typography to your Brand library through the Penpot API, and report any Penpot value that no longer matches the repo. Penpot’s built-in token support keeps this structured. Try: “Every night, sync colors from the repo’s tokens.json into our Brand library and list any Penpot values that don’t match.”
2. Keep a living Figma snapshot during migration
If you are mid-migration from Figma, “which version is current?” becomes a daily question. A scheduled export-and-import gives you a one-way answer: pull the Figma file on Sunday, import it into a Figma Mirror project, and diff the frame list against last week so everyone sees what changed. Be honest with the team that it is a snapshot, not a two-way sync — the value is knowing, not bidirectional magic. Try: “Each Sunday, refresh the Figma mirror project and post the list of added or removed frames to the design-migration channel.”
3. Ship new icons from Git into the library
Designers should not hand-carry SVGs from a repository into the icon library. The agent can watch the icons folder, and whenever a merge to main adds or changes an SVG, import it into your Icon Library file as a component named after the file, replacing the previous version. Your product code and your design tool stop quietly drifting apart, and nobody opens a ticket to copy a file. Try: “When an SVG merges into the icons folder on main, add it to the Icon Library as a component named after the file, and tell me if one gets removed.”
4. Publish a weekly design-system changelog
Teams ask “did the button change?” more often than you would think. Every Friday, the agent can diff your Brand library against last week’s version — new or renamed components, changed fills, removed assets — and post a plain-language summary to Slack or email. The changelog writes itself, and your designers stop answering the same question five times a week. Ask it to pin the summary to your team channel for a searchable history. Try: “Every Friday, diff the Brand library against last week and post added, changed, or removed components to the design channel.”
5. Route comments into Slack
Penpot comments are where feedback lives, but nobody watches the comment panel all day. Penpot can fire a webhook when a comment lands; the agent can receive it, look up who posted and on which file, and post a one-liner to your team channel with a link back. Feedback stops waiting for someone to notice, and threads get answered while the reviewer still remembers the context. Try: “Post to the website channel whenever someone comments on the Marketing Site file, with the author, the text, and a link.”
6. Send the sprint-end package to clients
The end-of-sprint ritual — export the screens, assemble a PDF, write the email — compresses into one scheduled job. The agent exports the pages your team tagged for client review as PDF plus PNGs, zips them, and emails the package to the stakeholder list with the project owner CC’d. Review the email draft the first time it runs, then let it carry itself; the format stays consistent, which clients quietly appreciate. Try: “Every other Friday, export pages tagged client-review as PDF and PNG, zip them, and email the package to the Acme stakeholders.”
7. Gate every release with a design lint
A release should never ship with an off-brand color or an unmarked asset. The agent can build a lint step that runs in CI or on a schedule: scan the production files, fail if a fill uses a hex outside the Brand library, if an icon is not marked for export, or if text sizes fall outside your scale. Catching this before release is cheap; a reprint or an emergency redeploy is not. Try: “Before each release, scan the production files and fail the build if any fill is not from the Brand library or any icon is not export-marked.”
8. Back up Penpot nightly — and prove the restore
Penpot keeps file data in Postgres and uploaded media on disk, so a real backup is both. The agent can schedule a nightly dump of the database plus the assets volume, push both to S3 with fourteen-day retention, and — the part everyone skips — restore last night’s backup into a throwaway container on the first of each month to prove it actually works. A backup you have never restored is a hypothesis. Try: “Back up Postgres and the assets volume to S3 nightly, keep fourteen days, and restore last night’s backup into a scratch instance on the first of each month — tell me if it fails.”
9. Turn a frame into a Tailwind draft
Developers prototype faster from real markup than from redlines. The agent can read a frame through the API — layout structure, text content, fills — and generate a rough HTML and Tailwind file into the repo’s drafts folder. It will not be production code, but it skips the blank-page hour and gives the pull request something concrete to react to. Try: “Convert the landing page hero frame into a Tailwind HTML draft in the repo and open a work-in-progress branch for it.”
10. Report which library components actually get used
Design systems accumulate components nobody uses, and nobody retires them because nobody counts. A weekly job can query the database for component instances across your files, rank library components by usage, and list the zeros — your retirement candidates. The decision stays human; the counting does not. Try: “Every Monday, rank Brand library components by how often they are instanced across files and list the ones nobody used in the last ninety days.”
See it in action
Official walkthroughs from the Penpot team:
Worth a watch next:
- Design tokens just got easier in Penpot 2.17 — token workflows are the backbone of the sync automations above.
- Comments in Penpot: feedback without leaving design mode — the comment flows you’ll be routing into Slack and email.
Each of these starts as a conversation and ends as a scheduled job with logs you can check. If you would rather start from a finished server than a compose file, see Penpot on OpenSysLab.
