Prompts and a skill that keep a help center in step with a product that ships every day: the written articles and their walkthrough videos. They're the generic versions of what Tella runs on every pull request; the story is in How I created self-updating help docs (with videos).
| File | What it is | How to run it |
|---|---|---|
docs-sync-prompt.md |
Reads each merged PR, updates the docs it affects, records or re-records the walkthrough video, opens a docs PR | An agent triggered on every merged PR (a GitHub Action on pull_request: closed + merged, or your agent platform's "PR merged" event) |
docs-drift-check-prompt.md |
Weekly audit: sidebar groups vs the real UI, stale labels, undocumented and unshipped features, and (optionally) whether article steps still work | An agent on a weekly schedule |
docs-skill/SKILL.md |
How your docs are organised and written: sidebar-mirrors-the-product, no URL moves, answer-first style, checks, handoff | Commit to your repo (e.g. .agents/skills/docs/) so every agent that touches docs reads it |
docs-skill/help-doc-video.md |
Recipe for making help videos with Tella Auto Capture: when to film, demo account, waiting for the deploy, the take, frame checks, styling, publishing and archiving | Committed next to the skill; the sync prompt follows it |
- Copy
docs-skill/into your repo and fill in its{{placeholders}}: docs paths, the commands that list your product's UI surfaces, your docs checks, and the reviewer. - Fill in the placeholders in the two prompts and set them up as automations. Delete the optional sections (Slack, videos, Auto Capture step checks) you don't want.
- For videos: connect your agent to the Tella MCP server (
https://api.tella.com/mcp), create a demo account in your product, save its login in Tella under Settings → Credentials, and write your house video style once (ideally as a Tella Agent workspace skill). Auto Capture needs a paid Tella plan and uses AI credits per take.
The agent never merges: every change arrives as a PR for a person to review.