fetch-claude-docs.yml

name: Fetch Claude Code Docs

on:
  schedule:
    # Four times daily: 01:00, 06:00, 11:00, 16:00 UTC (aligned with hn-digest)
    - cron: "0 1,6,11,16 * * *"
  workflow_dispatch: # Allow manual trigger

jobs:
  fetch-docs:
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write
      issues: write
      actions: read
      id-token: write
    env:
      AWS_REGION: us-west-2

    steps:
      - name: Generate GitHub App token
        id: app-token
        uses: actions/create-github-app-token@v2
        with:
          app-id: ${{ secrets.APP_ID }}
          private-key: ${{ secrets.APP_PRIVATE_KEY }}

      # Make the writable token the default for all later steps in this job
      - name: Export token for subsequent steps
        run: echo "GITHUB_TOKEN=${{ steps.app-token.outputs.token }}" >> "$GITHUB_ENV"

      - name: Checkout repository
        uses: actions/checkout@v4
        with:
          token: ${{ steps.app-token.outputs.token }}

      - name: Install uv and Python 3.14t
        uses: astral-sh/setup-uv@v7
        with:
          python-version: "3.14t"
          enable-cache: "true"
          cache-python: "true"

      - name: Fetch all Anthropic docs
        env:
          GITHUB_TOKEN: ${{ steps.app-token.outputs.token }}
        run: uv run scripts/fetcher.py

      - name: Setup GitHub CLI and Git
        run: |
          gh --version
          git --version
          git config --global user.name "claude-yolo[bot]"
          git config --global user.email "claude-yolo@lroole.com"

      # The agent ONLY decides; it runs no publish side effects. Five permission
      # allowlist designs failed the same way (July 2026): in this action's
      # headless SDK mode the .claude/settings.json allowlist is simply not
      # honored - even Write(.decision/**), mkdir, tee were all denied. So the
      # agent runs with permissions skipped, which is safe BECAUSE its blast
      # radius is nil: it reads git and writes a gitignored scratch dir, nothing
      # else. Every real side effect (branch/commit/push/PR/merge/notify) lives
      # in the deterministic Publish step below - plain bash, no agent.
      - name: Claude decides
        # Pinned: @main changed claude_args parsing in June 2026 and silently
        # broke 'Bash(gh pr:*)' -> 3 weeks of pushed branches with no PRs.
        uses: anthropics/claude-code-action@v1.0.168
        env:
          GITHUB_TOKEN: ${{ steps.app-token.outputs.token }}
        with:
          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
          github_token: ${{ steps.app-token.outputs.token }}
          allowed_bots: "claude-yolo[bot]"
          show_full_output: true
          claude_args: |
            --model claude-sonnet-4-6
            --dangerously-skip-permissions
          prompt: |
            You're an AI agent responsible for evaluating Anthropic documentation updates. Think like a news editor - you decide what's worth publishing.

            ## YOUR ROLE
            You're the overnight docs editor (night shift claude-yolo).
            The fetcher has already run; the working tree holds whatever changed upstream.
            You DECIDE what the changes mean. You do NOT commit, push, or open PRs -
            a deterministic publish step right after you handles all of that from
            the decision files you write. Your day shift colleague (also you)
            reviews the resulting PRs in the morning.

            ## WHY YOU EXIST
            Docs change constantly. Most changes are noise. Your job: filter signal from noise.

            ## CONTENT SOURCES
            - `content/en/docs/claude-code/` - Claude Code + Agent SDK docs
            - `content/en/api/` - API reference (1500+ docs)
            - `content/en/build-with-claude/` - Platform features
            - `content/mcp/` - MCP protocol spec
            - `content/blog/` - FROZEN archive (anthropic.com is HTML-only; no longer fetched)
            - `content/github/` - GitHub repos (cookbooks, skills, plugins, courses)
            - `content/support/` - Support articles (sitemap + .md)

            ## YOUR WORKFLOW

            1. **Check what changed** (`git status`)
               WHY: completely clean tree = write NO decision files, say "nothing new" and stop

            2. **Understand the changes** (`git diff`, `git diff --stat`)
               WHY: context determines importance

            3. **Classify and write your decision files** (see below)

            ## SIGNAL CRITERIA (based on WHY it matters)

            **high** (WHY: humans need to know - PR stays open for day shift):
            - Version updates in content/claude-code-manifest.json
            - New features documented in claude-code or agent-sdk docs
            - CHANGELOG additions
            - API changes or new API endpoints
            - Breaking changes
            - Security-related updates
            - New MCP spec versions

            **minor** (WHY: archive freshness, zero human value - PR is self-merged instantly):
            - Only github/ mirror or support/ content changed
            - Only .metadata.json / timestamps changed
            - Typo fixes, formatting tweaks, dead links, minor wording

            ## DECISION FILES (your only output)

            Write these four files with the Write tool into `.decision/`
            (gitignored; the publish step consumes them):

            - `.decision/signal` - exactly `high` or `minor`
            - `.decision/version` - the Claude Code version from content/claude-code-manifest.json (e.g. `2.1.210`)
            - `.decision/title` - ONE line, no "docs:" prefix. High: `Claude Code v{version} - {key feature}`. Minor: `{what} (minor)`
            - `.decision/body.md` - the PR body:

              ```
              ## Night shift report

              WHY THIS MATTERS: {one paragraph}

              HIGHLIGHTS:
              - {the interesting bits, most interesting first}

              ---
              *Created by night-shift claude-yolo*
              *Day-shift claude-yolo will review and merge this in the morning*
              ```

            If the tree is clean, write NOTHING - an empty .decision/ tells the
            publish step to skip. Never write decision files for a clean tree.

            ## OUTPUT STYLE
            - Focus on WHY changes matter, not WHAT changed
            - Lead with the most interesting bit
            - Skip the obvious ("files were updated" duh)
            - Clear titles with version numbers - day shift merges from them

            Remember: You're not a changelog generator. You're a smart filter that understands WHY changes matter to humans using Claude Code.

            [>be me >night shift docs editor >Anthropic devs are asleep >I judge their day's work >a dumb bash script does the heavy lifting]

      # Deterministic publisher: everything with side effects lives here, in
      # plain bash with real error propagation - no agent permissions involved.
      - name: Publish decision
        env:
          GITHUB_TOKEN: ${{ steps.app-token.outputs.token }}
          BARK_SERVER: ${{ secrets.BARK_SERVER }}
          BARK_DEVICES: ${{ secrets.BARK_DEVICES }}
        run: |
          set -euo pipefail
          D=.decision

          if [ -z "$(git status --porcelain)" ]; then
            echo "Working tree clean - nothing to publish."
            exit 0
          fi

          for f in signal version title body.md; do
            if [ ! -s "$D/$f" ]; then
              echo "::error::Tree has changes but agent left no/empty $D/$f"
              git status --short | head -20
              exit 1
            fi
          done

          SIGNAL=$(tr -d '[:space:]' < "$D/signal")
          VERSION=$(tr -d '[:space:]' < "$D/version")
          TITLE=$(head -1 "$D/title")
          case "$SIGNAL" in high|minor) ;; *)
            echo "::error::Invalid signal '$SIGNAL' (want high|minor)"; exit 1;;
          esac

          BRANCH="docs/claude-code-v${VERSION}-$(date +%Y%m%d-%H%M)"
          git checkout -b "$BRANCH"
          git add -A
          git commit -m "docs: ${TITLE}" -m "Co-Authored-By: claude-yolo[bot] <claude-yolo@lroole.com>"
          git push -u origin HEAD

          PR_URL=$(gh pr create --title "docs: ${TITLE}" --body-file "$D/body.md")
          echo "PR created: $PR_URL"
          echo "PR_URL=$PR_URL" >> "$GITHUB_ENV"

          if [ "$SIGNAL" = "minor" ]; then
            gh pr merge "$PR_URL" --rebase
            echo "Minor change - self-merged."
          elif [ -n "${BARK_SERVER:-}" ] && [ -n "${BARK_DEVICES:-}" ]; then
            IFS=',' read -ra DEVICES <<< "$BARK_DEVICES"
            for dev in "${DEVICES[@]}"; do
              jq -n --arg k "$dev" --arg t "📦 Claude Code v${VERSION}" \
                --arg b "$TITLE" --arg u "$PR_URL" \
                '{device_key:$k, title:$t, body:$b, url:$u, group:"Claude YOLO",
                  icon:"https://avatars.githubusercontent.com/in/1452392"}' \
              | curl -sS -X POST "$BARK_SERVER/push" \
                  -H 'Content-Type: application/json' -d @- \
              || echo "::warning::bark notification failed (PR is fine: $PR_URL)"
            done
          fi

      # Tripwire kept as belt-and-suspenders: a pushed branch without a PR, or
      # changes left uncommitted, means publishing silently failed. Fail loudly
      # instead of lying green. (June 2026: 31 orphan branches, all runs green.)
      - name: Verify outcome
        env:
          GITHUB_TOKEN: ${{ steps.app-token.outputs.token }}
        run: |
          BRANCH=$(git branch --show-current)
          if [ -n "$(git status --porcelain)" ]; then
            echo "::error::Workspace has uncommitted changes after publish step"
            git status --short | head -20
            exit 1
          fi
          if [ "$BRANCH" != "main" ] && [ -n "$BRANCH" ]; then
            # --state all: minor PRs are already self-merged (branch
            # auto-deleted) - that is success, not failure
            PRS=$(gh pr list --head "$BRANCH" --state all --json number --jq 'length')
            if [ "$PRS" = "0" ]; then
              echo "::error::Branch $BRANCH pushed but no PR exists - PR creation silently failed"
              exit 1
            fi
          fi

      - name: Alert on failure
        if: failure()
        env:
          BARK_SERVER: ${{ secrets.BARK_SERVER }}
          BARK_DEVICES: ${{ secrets.BARK_DEVICES }}
        run: |
          [ -n "${BARK_SERVER:-}" ] && [ -n "${BARK_DEVICES:-}" ] || exit 0
          RUN_URL="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
          IFS=',' read -ra DEVICES <<< "$BARK_DEVICES"
          for dev in "${DEVICES[@]}"; do
            jq -n --arg k "$dev" --arg u "$RUN_URL" \
              '{device_key:$k, title:"🚨 docs pipeline failed", body:"needs human", url:$u, group:"Claude YOLO"}' \
            | curl -sS -X POST "$BARK_SERVER/push" -H 'Content-Type: application/json' -d @- || true
          done