Prompt Clustering: add mandatory root-cause drilldown for low-merge outlier clusters

[Copilot]

The prompt-clustering workflow identified cluster-level merge-rate variance but did not require a blocker analysis for low-performing clusters. This left outlier clusters (notably image/container/digest-themed groups) without a concrete, evidence-backed root cause.

• Root-cause trigger for outlier clusters

  • Added explicit outlier detection criteria in clustering analysis:
    • cluster size threshold (>=15 PRs)
    • merge-rate gap threshold (>=10 percentage points below overall)
  • Requires focused investigation of the lowest-merge qualifying cluster.

• New investigation phase (Phase 4.5)

  • Added a mandatory blocker-classification pass over PR evidence (comments, reviews, reviewDecision, changed files).
  • Standardized blocker taxonomy:
    • CI/test failure (including container/image build failures)
    • review friction (including digest/security/compliance concerns)
    • scope/complexity mismatch
    • other/unknown
  • Requires evidence sampling (minimum 10 PRs when available), comparison against merged PRs in-cluster, and a primary-blocker conclusion with confidence.

• Report output upgrades

  • Added a dedicated Lowest-Merge Cluster Root Cause section to the report template.
  • Added blocker distribution table, merge-rate gap summary, evidence PR references, and targeted mitigation recommendations.
  • Updated success criteria to require this outlier investigation when applicable.

### Phase 4.5: Root-Cause Analysis for Low-Merge Clusters

When an outlier cluster is detected (>=15 PRs and merge rate >=10 points below overall rate), perform a focused blocker investigation for the lowest-merge cluster.
...
| Blocker Category | Count | Share | Notes |
|------------------|-------|-------|-------|
| CI/test failure  | X     | Y%    | ...   |
| Review friction  | X     | Y%    | ...   |

• Fixes [deep-report] Investigate why the image/container/digest Copilot PR cluster has the lowest merge rate (58.8% vs 74.3% fleet average) #51114

[Copilot]

Pull request overview

Adds mandatory root-cause analysis for statistically significant low-merge prompt clusters.

Changes:

• Defines outlier thresholds and blocker taxonomy.
• Adds evidence, comparison, and mitigation reporting requirements.
• Extends the report template and success criteria.

Show a summary per file

File	Description
.github/workflows/prompt-clustering-analysis.md	Adds outlier investigation and reporting instructions.
.github/workflows/prompt-clustering-analysis.lock.yml	Updates the compiled workflow body hash.

Review details

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Suppressed comments (1)

.github/workflows/prompt-clustering-analysis.md:595

• The investigation phase requires three concrete mitigation actions, while the report template requests only one outlier recommendation. Agents following the template can therefore omit two required mitigations. Provide three dedicated outlier mitigation slots.

4. **[Outlier Cluster Recommendation]**: [Action specifically targeting the lowest-merge cluster blocker]

• Files reviewed: 2/2 changed files
• Comments generated: 4
• Review effort level: Balanced

[github-actions]

🎉 This pull request is included in a new release.

Release: v0.86.1