feat: enable Claude Code Web visibility for automation workflows

claude · claude · commit 33138b2d4562 · 2025-11-24T12:15:17.000Z
Make automation workflows visible and trackable in Claude Code Web by using @claude triggers instead of direct API calls. This allows you to see the workflow progress live in your Claude Code Web session. Changes: 1. spec-to-code.yml (Code Generation): - Simplified to just trigger @claude with instructions - GitHub Action posts @claude comment when 'approved' label is added - Claude Code (via claude.yml) picks up the task - User sees generation progress in Claude Code Web - Claude Code creates implementations, commits, and makes PR 2. quality-check.yml (Quality Evaluation): - Posts @claude comment with evaluation instructions - Claude Code views preview images and evaluates - User sees evaluation progress in Claude Code Web - Claude Code posts results and adds labels 3. claude.yml (Core): - Enhanced permissions: contents:write, pull-requests:write, issues:write - Allows Claude Code to commit and create PRs - Fetch full history for better context Workflow Flow: 1. Issue gets 'approved' label 2. spec-to-code.yml posts @claude comment 3. claude.yml triggers (sees @claude) 4. Claude Code executes in Web (YOU SEE THIS LIVE!) 5. Claude Code commits implementations 6. Claude Code creates PR 7. test-and-preview.yml runs (regular CI) 8. quality-check.yml posts @claude comment 9. Claude Code evaluates (YOU SEE THIS LIVE!) 10. Claude Code posts quality report + labels Benefits: - 👁️ Full visibility in Claude Code Web - 🎮 Can intervene if needed - 📊 See progress in real-time - 🔍 Better debugging when things go wrong - 💬 Natural conversation-based workflow - 🤖 Same AI throughout (Claude Code Max)
diff --git a/.github/workflows/claude.yml b/.github/workflows/claude.yml
@@ -19,16 +19,16 @@ jobs:
       (github.event_name == 'issues' && (contains(github.event.issue.body, '@claude') || contains(github.event.issue.title, '@claude')))
     runs-on: ubuntu-latest
     permissions:
-      contents: read
-      pull-requests: read
-      issues: read
+      contents: write        # Allow commits
+      pull-requests: write   # Allow PR creation
+      issues: write         # Allow issue comments
       id-token: write
-      actions: read # Required for Claude to read CI results on PRs
+      actions: read         # Required for Claude to read CI results on PRs
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
         with:
-          fetch-depth: 1
+          fetch-depth: 0    # Full history for better context
 
       - name: Run Claude Code
         id: claude
diff --git a/.github/workflows/quality-check.yml b/.github/workflows/quality-check.yml
@@ -57,46 +57,64 @@ jobs:
           mkdir -p preview_images
           gsutil -m cp -r "gs://${{ steps.metadata.outputs.bucket }}/${{ steps.metadata.outputs.base_path }}/*" preview_images/ || echo "No images to download"
 
-      - name: Quality evaluation with Claude Code
-        uses: anthropics/claude-code-action@v1
+      - name: Trigger quality check with @claude
+        uses: actions/github-script@v7
         with:
-          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          prompt: |
-            TASK: Evaluate plot implementations against specifications
-
-            PR: #${{ steps.metadata.outputs.pr_number }}
-            Preview images location: preview_images/
-
-            Instructions:
-            1. List all PNG files in preview_images/ directory
-            2. For each preview image:
-               a. Parse filename to extract: spec_id, library, variant
-               b. Read corresponding spec file: specs/{spec_id}.md
-               c. View the preview image
-               d. Evaluate against quality criteria in spec
-
-            3. For each implementation, check:
-               - Does it meet ALL quality criteria listed in spec?
+          script: |
+            const prNumber = ${{ steps.metadata.outputs.pr_number }};
+            const bucket = '${{ steps.metadata.outputs.bucket }}';
+            const basePath = '${{ steps.metadata.outputs.base_path }}';
+
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: prNumber,
+              body: `@claude
+
+            ## 🎯 Task: Quality Evaluation
+
+            Preview images have been generated and uploaded to GCS. Please evaluate the implementations.
+
+            ### Instructions
+
+            1. **Preview images location**
+               - Already downloaded to \`preview_images/\` directory in this repository
+               - GCS Bucket: \`${bucket}\`
+               - GCS Path: \`${basePath}/\`
+
+            2. **For each preview image:**
+               - Parse filename to extract: spec_id, library, variant (format: \`{spec_id}_{library}_{variant}.png\`)
+               - Read corresponding spec file: \`specs/{spec_id}.md\`
+               - View the preview image carefully using your vision capabilities
+               - Evaluate against quality criteria listed in the spec
+
+            3. **Check each quality criterion:**
+               - Does it meet ALL quality criteria from the spec?
                - Are visual elements clear and readable?
                - Are colors appropriate and accessible?
                - Is the layout well-structured?
                - Score: 0-100 (≥85 to pass)
 
-            4. Generate quality report with:
-               - Overall verdict (PASS if all ≥85, FAIL otherwise)
-               - Score for each implementation
-               - Specific feedback for each quality criterion
+            4. **Generate quality report with:**
+               - Overall verdict (PASS if all implementations score ≥85, FAIL otherwise)
+               - Score for each implementation (matplotlib and seaborn separately)
+               - Specific feedback for each quality criterion (met/failed)
                - Strengths and improvements needed
+               - Concrete, actionable suggestions if FAIL
 
-            5. Use gh CLI to:
-               - Post quality report as comment on PR #${{ steps.metadata.outputs.pr_number }}
-               - Add label "quality-approved" if PASS
-               - Add label "quality-check-failed" if FAIL
+            5. **Update this PR:**
+               - Post quality report as comment on this PR
+               - Use \`gh pr edit\` to add appropriate label:
+                 - "quality-approved" if PASS
+                 - "quality-check-failed" if FAIL
 
-            Scoring Guidelines:
+            ### Scoring Guidelines
             - 90-100: Excellent - All criteria met, production ready
             - 85-89: Good - Minor issues, acceptable
-            - 75-84: Needs improvement
-            - <75: Rejected - Major issues
+            - 75-84: Needs improvement - regeneration recommended
+            - <75: Rejected - Major issues, regeneration required
+
+            Be objective and constructive in your feedback. Focus on measurable criteria.
 
-            Be objective and constructive in your feedback.
+            Preview images are ready - please proceed with the evaluation!`
+            });
diff --git a/.github/workflows/spec-to-code.yml b/.github/workflows/spec-to-code.yml
@@ -5,22 +5,14 @@ on:
     types: [labeled]
 
 jobs:
-  generate-code:
+  trigger-claude-code:
     # Only run when "approved" label is added
     if: github.event.label.name == 'approved'
     runs-on: ubuntu-latest
     permissions:
-      contents: write
-      pull-requests: write
       issues: write
-      id-token: write
 
     steps:
-      - name: Checkout repository
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
       - name: Extract spec ID from issue
         id: extract_spec
         env:
@@ -43,141 +35,62 @@ jobs:
           echo "spec_id=$SPEC_ID" >> $GITHUB_OUTPUT
           echo "✅ Extracted spec ID: $SPEC_ID"
 
-      - name: Check if spec file exists
-        id: check_spec
-        run: |
-          SPEC_ID="${{ steps.extract_spec.outputs.spec_id }}"
-          SPEC_FILE="specs/${SPEC_ID}.md"
-
-          if [ ! -f "$SPEC_FILE" ]; then
-            echo "Creating spec file from issue..."
-            mkdir -p specs
-            echo "${{ github.event.issue.body }}" > "$SPEC_FILE"
-
-            git config --global user.name "github-actions[bot]"
-            git config --global user.email "github-actions[bot]@users.noreply.github.com"
-            git add "$SPEC_FILE"
-            git commit -m "feat: add spec file for $SPEC_ID from issue #${{ github.event.issue.number }}"
-          else
-            echo "✅ Spec file exists: $SPEC_FILE"
-          fi
-
-      - name: Generate implementations with Claude Code
-        id: generate
-        uses: anthropics/claude-code-action@v1
+      - name: Trigger Claude Code with @claude comment
+        uses: actions/github-script@v7
         with:
-          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          prompt: |
-            TASK: Generate plot implementations from specification
+          script: |
+            const specId = '${{ steps.extract_spec.outputs.spec_id }}';
+
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+              body: `@claude
+
+            ## 🎯 Task: Generate Plot Implementations
 
-            SPEC_ID: ${{ steps.extract_spec.outputs.spec_id }}
-            ISSUE: #${{ github.event.issue.number }}
+            This issue has been approved! Please generate implementations for **${specId}**.
 
-            Instructions:
-            1. Read the spec file: specs/${{ steps.extract_spec.outputs.spec_id }}.md
-            2. Read generation rules: rules/generation/v1.0.0-draft/code-generation-rules.md
-            3. Read quality criteria: rules/generation/v1.0.0-draft/quality-criteria.md
-            4. Read self-review checklist: rules/generation/v1.0.0-draft/self-review-checklist.md
+            ### Instructions
 
-            5. Generate matplotlib implementation:
-               - Target: plots/matplotlib/scatter/${{ steps.extract_spec.outputs.spec_id }}/default.py
+            1. **Read the specification**
+               - File: \`specs/${specId}.md\`
+               - If file doesn't exist, create it from the issue body
+
+            2. **Read generation rules**
+               - \`rules/generation/v1.0.0-draft/code-generation-rules.md\`
+               - \`rules/generation/v1.0.0-draft/quality-criteria.md\`
+               - \`rules/generation/v1.0.0-draft/self-review-checklist.md\`
+
+            3. **Generate matplotlib implementation**
+               - Target: \`plots/matplotlib/scatter/${specId}/default.py\`
                - Follow generation rules exactly
                - Implement all spec requirements
                - Use type hints, docstrings, validation
-               - Add standalone execution block
+               - Add standalone execution block (\`if __name__ == '__main__':\`)
                - Self-review: Max 3 attempts to pass quality criteria
 
-            6. Generate seaborn implementation:
-               - Target: plots/seaborn/scatterplot/${{ steps.extract_spec.outputs.spec_id }}/default.py
+            4. **Generate seaborn implementation**
+               - Target: \`plots/seaborn/scatterplot/${specId}/default.py\`
                - Same requirements as matplotlib
 
-            7. Create summary in markdown format with:
-               - Libraries generated (matplotlib, seaborn)
-               - Self-review results for each
-               - File paths created
-               - Any issues encountered
+            5. **Create Pull Request**
+               - Branch: \`auto/${specId}\`
+               - Title: \`feat: implement ${specId}\`
+               - Include summary of generation results
+               - Link to this issue
 
-            IMPORTANT:
+            ### Requirements
             - Each implementation must pass self-review before moving to next
             - Use deterministic sample data (fixed seeds or hardcoded data)
             - Code must work standalone (no external files)
             - Follow Python 3.10+ syntax
+            - Commit with clear messages
 
-            Output your work summary when complete.
-
-      - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v5
-        with:
-          token: ${{ secrets.GITHUB_TOKEN }}
-          commit-message: |
-            feat: implement ${{ steps.extract_spec.outputs.spec_id }}
-
-            Auto-generated implementations from issue #${{ github.event.issue.number }}
-            Generated using Claude Code with versioned rules (v1.0.0-draft)
-
-            - matplotlib implementation
-            - seaborn implementation
-          branch: auto/${{ steps.extract_spec.outputs.spec_id }}
-          title: "feat: implement ${{ steps.extract_spec.outputs.spec_id }}"
-          body: |
-            ## 🤖 Auto-generated Implementation
-
-            This PR contains auto-generated code for **${{ steps.extract_spec.outputs.spec_id }}**
-
-            **Source:** Issue #${{ github.event.issue.number }}
-            **Generated by:** Claude Code Action with versioned rules (v1.0.0-draft)
-
-            ### What was generated
-
-            - ✅ matplotlib implementation (`plots/matplotlib/scatter/${{ steps.extract_spec.outputs.spec_id }}/default.py`)
-            - ✅ seaborn implementation (`plots/seaborn/scatterplot/${{ steps.extract_spec.outputs.spec_id }}/default.py`)
-
-            ### Next Steps
-
-            - [ ] Review generated code
-            - [ ] Run tests (triggered automatically on PR)
-            - [ ] Generate previews (triggered automatically on PR)
-            - [ ] Quality check (triggered automatically after previews)
-
-            ### Generation Process
-
-            1. ✅ Spec read from `specs/${{ steps.extract_spec.outputs.spec_id }}.md`
-            2. ✅ Generation rules applied from `rules/generation/v1.0.0-draft/`
-            3. ✅ Self-review loop completed (max 3 attempts per library)
-            4. ✅ Code committed to branch
-
-            ---
-
-            *Generated by [Claude Code Action](https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }})*
-          labels: |
-            code-generated
-            auto-pr
-          draft: false
-
-      - name: Comment on issue
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const specId = '${{ steps.extract_spec.outputs.spec_id }}';
-
-            await github.rest.issues.createComment({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              issue_number: context.issue.number,
-              body: `## 🤖 Code Generation Complete
-
-            Implementations for **${specId}** have been generated using Claude Code!
-
-            **Generated:**
-            - ✅ matplotlib
-            - ✅ seaborn
-
-            A pull request has been created with the generated code.
-
-            ### Next Steps
-            1. Multi-version tests will run automatically (Python 3.10-3.13)
-            2. Preview images will be generated
-            3. AI quality check will evaluate the implementations
+            ### Expected Outcome
+            - PR created with matplotlib and seaborn implementations
+            - Both pass self-review
+            - Tests will run automatically on PR
 
-            Check the [workflow run](https://github.com/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}) for details.`
+            Please proceed with the generation!`
             });
