Tina Content Audit Action

This repository provides a reusable GitHub Action workflow and helper scripts to audit content in a TinaCMS-powered repo. It queries your Tina content, runs AI feedback on selected files, opens issues with suggestions, and creates a PR to update the lastChecked timestamp.

What’s Included

• Workflow: audit-content.yml
  • Helper scripts (PNPM + TSX):
  • aggregate-links.ts
  • map-links-to-content.ts
  • update-checked.ts
• A Tina starter project with the GitHub action preconfigured
• A Python MCP tool for generating test blog posts: scripts/blog_server/main.py

How to Install

1. Copy the workflow file into your repo:

   • audit-content.yml

2. Copy the helper scripts into your repo under .github/scripts/:

   • tina-helpers

3. Commit and push.

4. In your repo Settings:

• Add Secrets:
  • TINA_CLIENT_ID
  • TINA_TOKEN
• Add Variables:
  • TINA_AUDITOR_COLLECTION,
  • TINA_AUDITOR_EXPIRY_DAYS,
  • TINA_AUDITOR_CONTENT_WINDOW,
  • optional
    • TINA_AUDITOR_SYSTEM_PROMPT

How to Use

• Go to Actions → “Audit Content with AI” → Run workflow.
• Merge any PRs that update lastChecked after reviewing issues.
  • the lastChecked timestamp must be updated to avoid re-auditing the same content.
• Assign any issues generated to GitHub copilot, or do them manually.

What It Does

• Aggregates content paths via Tina GraphQL
• Maps file paths to file contents
• Sends content to an AI model for feedback
• Creates GitHub issues per file with feedback
• Updates lastChecked frontmatter and opens a PR

Requirements

• A TinaCMS project with content files in your repo
• GitHub repository with Actions enabled
• GitHub models enabled for your repo/organization
• Secrets and variables set:
• Secrets:
  • TINA_CLIENT_ID
  • TINA_TOKEN
• Variables:
  • TINA_AUDITOR_COLLECTION (the name of the tina collection to audit)
  • TINA_AUDITOR_EXPIRY_DAYS (integer)
  • TINA_AUDITOR_CONTENT_WINDOW (integer: number of items to audit per run)
  • TINA_AUDITOR_SYSTEM_PROMPT (optional: system prompt for AI model)

How It Works

• Triggers manually via workflow_dispatch.

• Job: query-tina-content

  • Checks out the repo and sets up PNPM.
  • Writes Tina credentials and config into .github/scripts/tina-helpers/.env.
  • Runs helper scripts:
    • aggregate-links.ts reads TINA_AUDITOR_CONTENT_WINDOW and queries Tina to select a window of content paths.
    • map-links-to-content.ts reads each selected file and returns { path, content }.
  • Exposes two outputs:
    • items: array of { path, content } for the matrix.
    • links: array of file paths for later updates.

• Job: generate-feedback

  • Creates a matrix over items to process multiple files in parallel.
  • Logs a preview of each file’s content.
  • Sends content to the AI model with your TINA_AUDITOR_SYSTEM_PROMPT.
  • Creates a GitHub Issue per file with the AI feedback and labels it “🤖 Content Audit”.

• Job: update-checked

  • Re-checks out the repo and installs helper script deps.
  • Runs update-checked.ts with the links array to update lastChecked in frontmatter.
  • Formats the updated paths for display.
  • Opens a PR that updates lastChecked for the audited files, linking back to the workflow run.

Scripts

• A set of PNPM scripts to run in both GitHub and locally for content auditing.
• Location: .github/scripts/tina-helpers
• Run scripts:
• Local Scripts
  • pnpm run migrate-content - Adds a lastChecked property to all content files returned by the query in the .env file.
• Scripts for GitHub Action
  • pnpm run aggregate-links - Gets a set of content paths for a collection based on the query, expiry interval, and other settigns from the .env file.
  • pnpm run map-links-to-content <paths> - Takes a set of content paths and returns an array with their file contents.
  • pnpm run update-checked <paths> - Updates the lastChecked frontmatter for the given paths.

Notes

• Issues are labeled “🤖 Content Audit”.
• The PR branch is named content-audit/flow/<run_id>.
• The workflow is configured to run on workflow_dispatch (manual trigger).