chore(claude): add respond-to-discussion-features claude command (#15391)

# Overview

Adds a Claude command that automates finding GitHub discussions resolved
by recent releases and generates draft comments to inform the community.

Invoked with `/respond-to-discussion-features` command.

**NOTE:** This does not automatically respond. Only drafts a helpful
answer.

## Key Changes

- **New Claude command**:
`.claude/commands/respond-to-discussion-features.md`
- Fetches releases within a configurable time window (default: 2 months)
  - Parses release notes for `feat:` and `fix:` entries
  - Fetches all open Feature Request discussions via GraphQL
  - Matches features to discussions using keyword extraction
  - Generates markdown output with matches table and draft comments

## Design Decisions

The command is fully agentic—it provides structured instructions that
Claude executes step-by-step, using TodoWrite for progress tracking.
This approach allows flexibility in matching logic while maintaining
reproducibility through explicit process steps.

Output goes to `.claude/artifacts/` (gitignored) rather than requiring
manual cleanup of generated files.

## Overall Flow

```mermaid
sequenceDiagram
    participant User
    participant Claude
    participant GitHub API
    participant Output

    User->>Claude: /respond-to-discussion-features [time-period]
    Claude->>GitHub API: gh release list (get releases in window)
    Claude->>GitHub API: gh release view (fetch release notes)
    Claude->>Claude: Parse feat:/fix: entries, extract keywords
    Claude->>GitHub API: GraphQL paginate discussions
    Claude->>Claude: Match keywords to discussion titles/bodies
    Claude->>Output: Write .claude/artifacts/community-response-YYYY-MM-DD.md
    Claude->>User: "Found N matches. Output written to..."
```

Author: denolfe

.claude/commands/respond-to-discussion-features.md

@@ -0,0 +1,135 @@
+---
+description: Find GitHub discussions resolved by recent releases and draft response comments to inform the community.
+argument-hint: <time-period> (default: 2months, e.g., "3months", "6weeks", "2025-11-01")
+allowed-tools: Bash(gh release list:*, gh release view:*, gh api graphql:*), Glob, Grep, Read, Write, TodoWrite
+---
+
+# Respond to Community
+
+Find open GitHub discussions in "Feature Requests & Ideas" that have been resolved by recent releases, and generate draft comments to post.
+
+## Arguments
+
+- `$ARGUMENTS` - Time period to analyze (default: `2months`)
+- Accepts: `Ndays`, `Nweeks`, `Nmonths`, or ISO date `YYYY-MM-DD`
+
+## Process
+
+### Step 1: Setup & Fetch Releases
+
+1. Parse time period from `$ARGUMENTS` (default: 2months)
+2. Create TodoWrite with: Fetch releases | Parse release notes | Fetch discussions | Match features to discussions | Generate output
+3. Run `gh release list --repo payloadcms/payload --limit 30` to get recent releases
+4. Filter to releases within the time period
+5. For each release, fetch notes: `gh release view <tag> --repo payloadcms/payload --json body -q '.body'`
+
+### Step 2: Parse Release Notes
+
+1. Extract feature and fix entries from each release:
+   - Lines containing `feat:` or `feat(`
+   - Lines containing `fix:` or `fix(`
+2. For each entry, capture:
+   - PR title (the description text)
+   - PR number (from `#NNNN` pattern)
+   - Package scope if present (from `feat(scope):` pattern)
+   - Version it shipped in
+3. Build keyword list from titles (feature names, package names, key terms)
+
+### Step 3: Fetch Discussions
+
+1. Use GraphQL to paginate through all open Feature Request discussions:
+
+```
+gh api graphql -f query='
+{
+  repository(owner: "payloadcms", name: "payload") {
+    discussions(first: 100, states: OPEN, categoryId: "MDE4OkRpc2N1c3Npb25DYXRlZ29yeTMyMzY4NTUx", orderBy: {field: CREATED_AT, direction: DESC}) {
+      pageInfo { hasNextPage endCursor }
+      nodes {
+        number
+        title
+        url
+        createdAt
+        body
+      }
+    }
+  }
+}'
+```
+
+2. Continue paginating with `after: "<endCursor>"` until all discussions fetched
+3. Store: number, title, URL, createdAt, body (for matching)
+
+### Step 4: Match Features to Discussions
+
+For each feature/fix from releases:
+
+1. Search discussion titles and bodies for keyword matches
+2. Consider semantic variations (e.g., "groupBy" matches "group by", "group-by")
+3. Filter: only include if discussion was created BEFORE the feature's release date
+4. Score matches by keyword overlap
+5. Flag ambiguous matches as "Needs Verification"
+
+Multiple discussions can match the same PR - include all matches.
+
+### Step 5: Generate Output
+
+1. Create output directory if needed: `.claude/artifacts/`
+2. Generate markdown file: `.claude/artifacts/community-response-YYYY-MM-DD.md`
+
+**Output structure:**
+
+```markdown
+# Community Response: Features Shipped [start date] → [end date]
+
+## Summary
+
+Found **N discussions** resolved by releases vX.XX.X → vX.XX.X
+
+## Matches
+
+| Discussion Title | Discussion Link                                        | PR Title   | PR Link                                         |
+| ---------------- | ------------------------------------------------------ | ---------- | ----------------------------------------------- |
+| [title]          | https://github.com/payloadcms/payload/discussions/NNNN | [pr title] | https://github.com/payloadcms/payload/pull/NNNN |
+
+## Draft Comments
+
+### 1. [Discussion Title]
+
+**Link:** https://github.com/payloadcms/payload/discussions/NNNN
+
+> [Draft comment]
+
+---
+```
+
+**Comment template:**
+
+- Lead with "Shipped in vX.XX.X 🎉" or "Fixed in vX.XX.X 🎉"
+- Brief description of what shipped (1-2 sentences)
+- Code example if the feature involves configuration
+- PR link at the end
+- Keep concise: 3-6 lines typical
+
+3. Present summary to user: "Found N matches. Output written to .claude/artifacts/community-response-YYYY-MM-DD.md"
+
+## Edge Cases
+
+| Scenario                             | Action                                        |
+| ------------------------------------ | --------------------------------------------- |
+| No `$ARGUMENTS`                      | Use default: 2months                          |
+| No releases in time period           | Report "No releases found" and stop           |
+| No matches found                     | Create output with "No matches found" message |
+| API rate limited                     | Report error, suggest waiting and retrying    |
+| Ambiguous match                      | Include with "⚠️ Needs Verification" note     |
+| Discussion already has team response | Include anyway - human reviewer will skip     |
+
+## Remember
+
+- Create TodoWrite with 5 steps upfront
+- Mark steps in_progress before starting, completed after finishing
+- Full URLs in output table (not just issue numbers)
+- Comments should be helpful and concise
+- Better to surface a false positive than miss a real match
+- Human will review before posting - don't need to be perfect
+- Auto-save to `.claude/artifacts/` directory