Merge pull request ag-grid#13321 from ag-grid/sync/sync-ag-shared

Sync ag-shared from ag-charts sync-ag-shared

Author: alantreadway

.rulesync/README.md

@@ -55,19 +55,21 @@ Quick-reference for all AI agent commands, skills, sub-agents, and rules availab
 
 ## Planning and Analysis
 
-| Type  | Name                            | Invoke                               | What it does                                     |
-| ----- | ------------------------------- | ------------------------------------ | ------------------------------------------------ |
-| Skill | 🔵 `plan-review`                | `/plan-review` (user)                | Review plans for completeness and correctness    |
-| Skill | 🔵 `plan-implementation-review` | `/plan-implementation-review` (user) | Review plan execution, identify delivery gaps    |
-| Skill | 🔵 `jira`                       | `/jira`                              | Create, estimate, or analyse JIRA tickets        |
-| Agent | 🔵 `nx-expert`                  | Auto                                 | Nx monorepo configuration and build optimisation |
+| Type  | Name                            | Invoke                               | What it does                                         |
+| ----- | ------------------------------- | ------------------------------------ | ---------------------------------------------------- |
+| Skill | 🔵 `jira`                       | `/jira`                              | Create, estimate, or analyse JIRA tickets            |
+| Skill | 🔵 `nx-performance`             | `/nx-performance`                    | Nx monorepo performance diagnostics and optimization |
+| Skill | 🔵 `plan-implementation-review` | `/plan-implementation-review` (user) | Review plan execution, identify delivery gaps        |
+| Skill | 🔵 `plan-review`                | `/plan-review` (user)                | Review plans for completeness and correctness        |
+| Agent | 🔵 `nx-expert`                  | Auto                                 | Nx monorepo configuration and build optimisation     |
 
 ## Prompt Hygiene
 
 | Type  | Name                  | Invoke                     | What it does                                     |
 | ----- | --------------------- | -------------------------- | ------------------------------------------------ |
-| Skill | 🔵 `validate-prompts` | `/validate-prompts` (user) | Validate prompt file references for path hygiene |
+| Skill | 🔵 `reflect`          | `/reflect` (user)          | Self-reflection and meta-cognitive analysis      |
 | Skill | 🔵 `rulesync`         | `/rulesync`                | Configure AI/agentic tooling via `.rulesync/`    |
+| Skill | 🔵 `validate-prompts` | `/validate-prompts` (user) | Validate prompt file references for path hygiene |
 
 ## Memory
 
@@ -144,12 +146,14 @@ Skills load on-demand when invoked. All skills are invoked via `/skill-name`. Al
 | 🔵 `git-split`                  | Split large files preserving git history                    |
 | 🔵 `git-worktree-clean`         | Hard-reset worktree to `origin/latest`                      |
 | 🔵 `jira`                       | Create, estimate, or analyse JIRA tickets                   |
+| 🔵 `nx-performance`             | Nx monorepo performance diagnostics and optimization        |
 | 🔵 `plan-implementation-review` | Review plan execution, identify delivery gaps               |
 | 🔵 `plan-review`                | Review plans for completeness and correctness               |
 | 🔵 `pr-create`                  | Commit, push, and open a PR                                 |
 | 🔵 `pr-review`                  | Review a PR (Markdown default, JSON with `--json`)          |
 | 🔵 `pr-split`                   | Split a branch into stacked PRs                             |
 | 🔵 `recall`                     | Load branch context, browse project memories                |
+| 🔵 `reflect`                    | Self-reflection and meta-cognitive analysis                 |
 | 🔵 `remember`                   | Save branch context or project learnings as memory          |
 | 🔵 `rulesync`                   | Configure AI/agentic tooling via `.rulesync/`               |
 | 🔵 `sync-ag-shared`             | Sync ag-shared subrepo changes across AG repos              |

.rulesync/skills/nx-performance

@@ -0,0 +1 @@
+../../external/ag-shared/prompts/skills/nx-performance/

.rulesync/skills/reflect

@@ -0,0 +1 @@
+../../external/ag-shared/prompts/skills/reflect/

external/ag-shared/.gitrepo

@@ -6,7 +6,7 @@
 [subrepo]
 	remote = https://github.com/ag-grid/ag-shared.git
 	branch = latest
-	commit = 8ba8e8e436eed4992181a97843b045a0c5249aef
+	commit = 2459687b24688dc840629ad808471e7ad7945a8f
 	parent = d0d058059e38acc51bd4484e3861c57d6cada8b7
 	method = rebase
 	cmdver = 0.4.9

external/ag-shared/docs/SYNC-LOG.md

@@ -6,6 +6,25 @@ Newest entries first. Generated by `/ag-shared-sync-log`.
 
 ---
 
+## 2026-03-17 -- Migrate nx-performance skill to ag-shared
+
+**Branch:** `latest`
+
+### Changes
+
+- **[prompts/skills]** Migrated `nx-performance` skill from ag-charts-local to ag-shared — generalized AG Charts-specific language to "AG product monorepos" while preserving concrete examples
+- **[prompts/skills]** Skill structure: `SKILL.md` + `references/` subdirectory with `caching-strategy.md`, `build-and-dev.md`, `ci-patterns.md`, `gotchas.md`
+
+### Migration Actions
+
+- [ ] ag-charts: Replace `.rulesync/skills/nx-performance/` directory with symlink → `../../external/ag-shared/prompts/skills/nx-performance/`
+- [ ] ag-grid: Add `.rulesync/skills/nx-performance` symlink → `../../external/ag-shared/prompts/skills/nx-performance/`
+- [ ] ag-studio: Add `.rulesync/skills/nx-performance` symlink → `../../external/ag-shared/prompts/skills/nx-performance/`
+- [ ] Run `./external/ag-shared/scripts/setup-prompts/setup-prompts.sh`
+- [ ] Run `./external/ag-shared/scripts/setup-prompts/verify-rulesync.sh`
+
+---
+
 ## 2026-03-12 -- Context optimisation, plunker multi-product, symlinked repo support
 
 **Branch:** `latest`

external/ag-shared/github/actions/codex-pr-review/action.yml

@@ -288,6 +288,93 @@ runs:
               echo "prompt_file=${prompt_file}" >> $GITHUB_OUTPUT
               echo "output_ext=${output_ext}" >> $GITHUB_OUTPUT
 
+        - name: Generate PR Diff Context
+          if: steps.check-auth.outputs.skipped != 'true'
+          id: diff-context
+          shell: bash
+          env:
+              PR_NUMBER: ${{ inputs.pr-number }}
+              BASE_REF: ${{ inputs.base-ref }}
+              HEAD_REF: ${{ inputs.head-ref }}
+              PR_TITLE: ${{ inputs.pr-title }}
+              PR_AUTHOR: ${{ inputs.pr-author }}
+              ARGUMENTS: ${{ inputs.inline-comments == 'true' && format('--json {0}', inputs.pr-number) || inputs.pr-number }}
+              GH_TOKEN: ${{ inputs.github-token }}
+          run: |
+              # Pre-generate the diff and metadata so Codex doesn't need to exec git commands.
+              # This avoids sandbox (bwrap) issues on GitHub Actions runners.
+              context_file=".codex-diff-context-${PR_NUMBER}.md"
+              diff_file=".codex-diff-${PR_NUMBER}.patch"
+              echo "context_file=${context_file}" >> $GITHUB_OUTPUT
+              echo "diff_file=${diff_file}" >> $GITHUB_OUTPUT
+
+              PR_REF="origin/pr/${PR_NUMBER}"
+              DIFF_BASE="origin/${BASE_REF}...${PR_REF}"
+              HEAD_SHA=$(git rev-parse "${PR_REF}")
+              PR_URL="https://github.com/${GITHUB_REPOSITORY}/pull/${PR_NUMBER}"
+
+              # Fill in missing metadata (e.g. workflow_dispatch doesn't have PR event context)
+              if [ -z "${PR_TITLE}" ] || [ -z "${PR_AUTHOR}" ] || [ -z "${HEAD_REF}" ]; then
+                PR_META=$(gh pr view "${PR_NUMBER}" --json title,author,headRefName 2>/dev/null || echo '{}')
+                [ -z "${PR_TITLE}" ] && PR_TITLE=$(echo "${PR_META}" | jq -r '.title // empty')
+                [ -z "${PR_AUTHOR}" ] && PR_AUTHOR=$(echo "${PR_META}" | jq -r '.author.login // empty')
+                [ -z "${HEAD_REF}" ] && HEAD_REF=$(echo "${PR_META}" | jq -r '.headRefName // empty')
+              fi
+
+              # Collect diff stats
+              FILES_CHANGED=$(git diff --name-only "${DIFF_BASE}" | wc -l | tr -d ' ')
+              DIFF_STAT=$(git diff --stat "${DIFF_BASE}")
+              LINES_ADDED=$(git diff --numstat "${DIFF_BASE}" | awk '{ s += $1 } END { print s+0 }')
+              LINES_REMOVED=$(git diff --numstat "${DIFF_BASE}" | awk '{ s += $2 } END { print s+0 }')
+
+              # Write the raw diff to a standalone file (avoids Markdown escaping issues)
+              git diff "${DIFF_BASE}" > "${diff_file}"
+
+              # Build the context file (metadata only, no embedded diff).
+              # Uses printf instead of a heredoc to keep content indented within the
+              # YAML block scalar — unindented heredoc lines break GitHub Actions' YAML parser.
+              COMMIT_MSGS=$(git log --format="%s" "origin/${BASE_REF}..${PR_REF}")
+              printf '%s\n' \
+                "" \
+                "## Pre-generated PR Context" \
+                "" \
+                "The PR diff and metadata have been pre-generated and are provided below." \
+                "**Do NOT run any git, gh, or shell commands.** Analyse the diff below directly." \
+                "" \
+                "### Arguments" \
+                "" \
+                "ARGUMENTS: ${ARGUMENTS}" \
+                "" \
+                "### PR Metadata" \
+                "- **PR Number:** ${PR_NUMBER}" \
+                "- **PR Title:** ${PR_TITLE}" \
+                "- **PR URL:** ${PR_URL}" \
+                "- **Author:** ${PR_AUTHOR}" \
+                "- **Base Branch:** ${BASE_REF}" \
+                "- **Head Branch:** ${HEAD_REF}" \
+                "- **Head SHA:** ${HEAD_SHA}" \
+                "- **Files Changed:** ${FILES_CHANGED}" \
+                "- **Lines Added:** ${LINES_ADDED}" \
+                "- **Lines Removed:** ${LINES_REMOVED}" \
+                "" \
+                "### Diff Stats" \
+                '```' \
+                "${DIFF_STAT}" \
+                '```' \
+                "" \
+                "### Commit Messages" \
+                '```' \
+                "${COMMIT_MSGS}" \
+                '```' \
+                "" \
+                "### Full Diff" \
+                "" \
+                "Everything below this line is the raw unified diff for this PR." \
+                "---" \
+                > "${context_file}"
+
+              echo "Generated context ($(wc -c < "${context_file}") bytes) + diff ($(wc -c < "${diff_file}") bytes)"
+
         - name: Run Codex Review
           if: steps.check-auth.outputs.skipped != 'true'
           id: codex
@@ -306,15 +393,17 @@ runs:
               # Output file paths (unique per PR to avoid conflicts)
               output_file=".codex-review-${{ inputs.pr-number }}.${{ steps.prompt.outputs.output_ext }}"
               error_file=".codex-error-${{ inputs.pr-number }}.log"
+              context_file="${{ steps.diff-context.outputs.context_file }}"
+              diff_file="${{ steps.diff-context.outputs.diff_file }}"
 
               echo "output_file=${output_file}" >> $GITHUB_OUTPUT
 
               # Run Codex with ALL output suppressed
-              # - stdin: Prompt file content piped in
+              # - stdin: Review instructions + metadata context + raw diff (three files concatenated)
               # - stdout (>/dev/null): Suppresses reasoning/streaming output
               # - stderr (2>file): Captured to file for debugging, not logged
               set +e
-              cat "${{ steps.prompt.outputs.prompt_file }}" | codex exec \
+              cat "${{ steps.prompt.outputs.prompt_file }}" "${context_file}" "${diff_file}" | codex exec \
                 --output-last-message "${output_file}" \
                 --sandbox read-only \
                 2>"${error_file}" \
@@ -672,4 +761,4 @@ runs:
           if: always()
           shell: bash
           run: |
-              rm -f ".codex-review-${{ inputs.pr-number }}.md" ".codex-review-${{ inputs.pr-number }}.json" ".codex-error-${{ inputs.pr-number }}.log"
+              rm -f ".codex-review-${{ inputs.pr-number }}.md" ".codex-review-${{ inputs.pr-number }}.json" ".codex-error-${{ inputs.pr-number }}.log" ".codex-diff-context-${{ inputs.pr-number }}.md" ".codex-diff-${{ inputs.pr-number }}.patch"

external/ag-shared/prompts/skills/jira/SKILL.md

@@ -23,6 +23,21 @@ context: fork
 
 Unified skill for creating, estimating, and analysing JIRA tickets across AG products.
 
+## Step 0: Verify Atlassian MCP Connection
+
+Before doing anything else, verify the Atlassian MCP is available by calling `mcp__atlassian__atlassianUserInfo`. If the tool is not available, returns an error, or the MCP server is not connected:
+
+**Hard stop.** Output the following and do not proceed:
+
+```
+Cannot proceed — Atlassian MCP is not connected.
+
+This skill requires the Atlassian MCP server to interact with JIRA.
+Please ensure the MCP connection is configured and active, then retry.
+```
+
+Do not attempt to work around the missing connection or produce partial results.
+
 ## Product Detection
 
 Detect the product from the repository context: