fix(prompt-clustering): add explicit safeoutputs create_discussion call

.github/workflows/prompt-clustering-analysis.md

@@ -565,6 +565,22 @@ Based on clustering analysis:
 _Generated by Prompt Clustering Analysis (Run: [run_id])_
 ```
 
+**After writing the report, post it as a discussion.**
+
+Write the fully-rendered report body to `/tmp/gh-aw/agent/discussion_body.md`, then build the safeoutputs payload and post it via stdin:
+
+```bash
+jq -n \
+  --arg title "Prompt Clustering Analysis - $(date +%Y-%m-%d)" \
+  --rawfile body /tmp/gh-aw/agent/discussion_body.md \
+  '{title: $title, body: $body}' \
+  > /tmp/gh-aw/agent/create_discussion_payload.json
+
+safeoutputs create_discussion . < /tmp/gh-aw/agent/create_discussion_payload.json
+```
+
+If `create_discussion` fails, call `noop` with a brief explanation so the failure is recorded.
+
 ### Phase 6: Cache Management
 
 Use the cache to avoid re-analyzing the same PRs:
@@ -641,4 +657,4 @@ If workflow logs unavailable for most PRs:
 - Note limitation in report
 - Focus on prompt analysis without turn counts
 
-Now analyze the prompts and generate your comprehensive report!
+Now analyze the prompts and generate your comprehensive report. When the analysis is complete, post it using `safeoutputs create_discussion` as described above. If you cannot complete the analysis, call `noop` with an explanation.

cmd/gh-aw/format_list_test.go

@@ -46,7 +46,6 @@ func TestFormatListWithOr(t *testing.T) {
 	}
 
 	for _, tt := range tests {
-		tt := tt
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 			result := formatListWithOr(tt.items)