[pull] main from github:main

.github/aw/report.md

@@ -40,23 +40,18 @@ safe-outputs:
 
 ### Header Levels
 
-**Use `###` or lower for all headers in your report to maintain proper document hierarchy.**
-
-When creating GitHub issues or discussions:
-- Use `###` (h3) for main sections (e.g., `### Test Summary`)
-- Use `####` (h4) for subsections (e.g., `#### Device-Specific Results`)
-- Never use `##` (h2) or `#` (h1) in reports — these are reserved for titles
+- Use `###` (h3) for main sections — e.g., `### Test Summary`
+- Use `####` (h4) for subsections — e.g., `#### Device-Specific Results`
+- Never use `##` (h2) or `#` (h1) — those are reserved for titles
 
 ### Progressive Disclosure
 
-**Wrap detailed content in `<details><summary>Section Name</summary>` tags to improve readability and reduce scrolling.**
-
-Use collapsible sections for:
+Wrap detailed content in `<details><summary>Section Name</summary>` tags. Use for:
 - Verbose details (full logs, raw data)
 - Secondary information (minor warnings, extra context)
 - Per-item breakdowns when there are many items
 
-Always keep critical information visible (summary, critical issues, key metrics).
+Keep critical information visible (summary, critical issues, key metrics).
 
 ### Report Structure Pattern
 
@@ -65,14 +60,6 @@ Always keep critical information visible (summary, critical issues, key metrics)
 3. **Details**: Use `<details><summary>Section Name</summary>` for expanded content
 4. **Context**: Add helpful metadata (workflow run, date, trigger)
 
-### Design Principles
-
-Reports should:
-- **Build trust through clarity**: Most important info immediately visible
-- **Exceed expectations**: Add helpful context like trends, comparisons
-- **Create delight**: Use progressive disclosure to reduce overwhelm
-- **Maintain consistency**: Follow patterns across all reports
-
 ### Example Report Structure
 
 ```markdown

.github/workflows/auto-triage-issues.md

@@ -53,6 +53,7 @@ safe-outputs:
     category: "audits"
     close-older-discussions: true
     max: 1
+  noop:
 timeout-minutes: 15
 features:
   copilot-requests: true
@@ -97,9 +98,13 @@ When running on schedule:
 When triggered manually as a backfill pass:
 
 1. **Fetch ALL open issues without any labels** using GitHub tools — do not limit to a fixed count
-2. **Process up to 10 unlabeled issues** in this run (respecting safe-output limits); if more exist, note the remainder in the report
-3. **Apply labels** to each issue based on classification rules below, using title/body heuristics and existing triage rules
-4. **Create a summary report** as a discussion listing every issue processed, the labels applied, and how many unlabeled issues (if any) still remain for the next pass
+2. **If there are no unlabeled issues**, call `noop` with "No unlabeled issues found during manual backfill — no action needed" and stop. Do not create a discussion.
+
+When unlabeled issues exist:
+
+3. **Process up to 10 unlabeled issues** in this run (respecting safe-output limits); if more exist, note the remainder in the report
+4. **Apply labels** to each issue based on classification rules below, using title/body heuristics and existing triage rules
+5. **Create a summary report** as a discussion listing every issue processed, the labels applied, and how many unlabeled issues (if any) still remain for the next pass
 
 ## Classification Rules
 
@@ -287,6 +292,16 @@ When running on schedule, create a discussion report following these formatting
 - **Learn from patterns** - Over time, notice which types of issues are frequently unlabeled
 - **Human override** - Maintainers can change labels; this is automation assistance, not replacement
 
+## Mandatory Completion Rule
+
+**Before finishing, check whether you called any safe-output tool in this run.** If you did NOT call `add_labels` or `create_discussion`, you MUST call `noop`. Every run MUST end with at least one safe-output call — failing to do so causes the workflow to fail with a safe-output compliance error.
+
+Situations that require a `noop` call:
+- No unlabeled issues were found (all trigger types)
+- The triggering issue already had appropriate labels
+- All issues analyzed were already labeled or had been processed
+- You were uncertain and chose not to label rather than guess incorrectly
+
 ## Success Metrics
 
 - Reduce unlabeled issue percentage from 8.6% to <5%

.github/workflows/copilot-pr-nlp-analysis.md

@@ -92,6 +92,7 @@ Generate a daily NLP-based analysis report of Copilot-created PRs merged within
 
 - Python analysis dependencies are already installed by pre-agent workflow steps.
 - **Do NOT run any `pip install` commands in agent turns.**
+- If an import unexpectedly fails, report the missing package in the output and continue with reduced analysis instead of installing dependencies in agent turns.
 - Run Python scripts with `/tmp/gh-aw/venv/bin/python3` to use the preinstalled environment.
 
 ## Task Overview