[prompt-clustering] Prompt Clustering Analysis - 2026-10-28

[github-actions[bot]]

Summary

Analysis Period: Last 30 days (2026-09-28 to 2026-10-27)
Total Tasks Analyzed: 605
Clusters Identified: 7
Overall Success Rate: 84.0% (508 merged / 96 closed / 1 open)

Full Analysis Report

General Insights

• Most Common Task Type: C1 — Agent/Workflow Orchestration & GitHub Integration (211 PRs, 35% of total)
• Highest Success Rate: C6 — PR Sous Chef Bot-Driven Fixes (89.4% merged)
• Most Complex Tasks: C5 — Dependency & Image Version Pinning (avg. 164.5 files changed per PR — large, mechanical bulk-pin diffs) and C6 — PR Sous Chef Bot-Driven Fixes (avg. 13.5 comments, 6.3 reviews — most discussion-heavy, via bot-driven iteration)
• Outliers: None flagged this run. Every cluster's merge rate (81.7%–89.4%) sits within ~8 points of the 84.0% overall rate, and none meets the outlier bar (cluster size ≥ 15 PRs and merge rate ≥ 10 points below overall). This is a change from the 2026-10-26 report, which flagged a 63.5%-merge "gh-aw Compiler/CLI Core & Infra Hardening" cluster as an outlier (16.2 points below that run's 79.7% overall rate) — see Key Findings.

Methodology note: clustering used TF-IDF (1-2 grams, 300 features, min document frequency 2) + K-means over cleaned PR title/body text, with k selected by silhouette score across k=3–8 (best: k=7, silhouette 0.054 — a fairly flat, mixed-topic score consistent with this repo's prompts sharing a lot of common vocabulary across task types). Full PR data (comments, reviews, commits, file-change counts) was available in cache for all 605 PRs in the window, so per-cluster engagement stats below reflect complete coverage, not a sample. No workflow-run turn/duration/cost data is included — see Limitations.

Cluster Analysis

Cluster 1: Agent/Workflow Orchestration & GitHub Integration

• Size: 211 tasks (34.9% of total — largest cluster)
• Success Rate: 82.0% (173 merged / 38 closed / 0 open)
• Avg Comments / Reviews / Files Changed: 5.5 / 3.7 / 46.3
• Top Keywords: agent, workflow, add, github, repository, coverage, workflows, job
• Characteristics: Broadest "glue" cluster — GitHub Actions job/workflow orchestration, Copilot agent job plumbing, firewall/egress scoping, checkout and bootstrap handling. Largest cluster by far yet still converges reliably despite topic diversity.
• Example PRs: Allow on.bots Apps to check out same-repository PR edits #72466, Allow on.bots Apps to check out same-repository PRs on edited events #72403, Separate Agy bootstrap downloads from runtime firewall egress #72102, Narrow Chrome egress to browser-specific Google services #72101, Deduplicate Copilot agent cleanup logic #72097

---

Cluster 2: Model Routing (Copilot/Claude/Pi)

• Size: 93 tasks (15.4% of total)
• Success Rate: 86.0% (80 merged / 12 closed / 1 open)
• Avg Comments / Reviews / Files Changed: 6.6 / 4.7 / 45.8
• Top Keywords: model, routing, copilot, claude, models, routed, smoke, agent
• Characteristics: Engine/model-routing logic — which AI engine and model a workflow runs under. Holds the dataset's only remaining open PR (Support Claude on Vertex AI with GCP WIF #72273, "Support Claude on Vertex AI with GCP WIF").
• Example PRs: Exclude Copilot utility models from alias selection and restore detection fallbacks #72404, Support Claude on Vertex AI with GCP WIF #72273, Replace unavailable Copilot model in video analyzer #71620, Prevent arXiv researcher Codex startup failure #71608, Attribute Pi main/sub-agent requests with x-awf-caller and preserve caller metadata #71043

---

Cluster 3: Safe-Output Failure Handling & Diagnostics

• Size: 86 tasks (14.2% of total)
• Success Rate: 83.7% (72 merged / 14 closed / 0 open)
• Avg Comments / Reviews / Files Changed: 5.6 / 4.9 / 44.5
• Top Keywords: safe, output, safe output, failure, issue, issues, outputs, diagnostics
• Characteristics: Safe-output transport reliability, failure-mode diagnostics, and unified-session accounting for job/script outcomes.
• Example PRs: Suppress startup-failure reports for cancelled agent jobs #72405, Record safe job and script outcomes in unified sessions #72146, Record privacy-preserving safe-output outcomes in unified sessions #72138, Enforce ubuntu-slim timeout limits and upgrade default runners for longer jobs #72131, Detect parenthesized Flush discards in uncheckedflushreturn #71896

---

Cluster 4: Ledger/Memory & Test Coverage Fixes

• Size: 71 tasks (11.7% of total)
• Success Rate: 81.7% (58 merged / 13 closed / 0 open) — lowest merge rate this cycle, but only 2.3 points below overall
• Avg Comments / Reviews / Files Changed: 4.8 / 2.9 / 13.5 (smallest average diffs of any cluster)
• Top Keywords: ledger, memory, test, remove, validation, coverage, findings, fixes
• Characteristics: Repo-memory/ledger primitives, targeted test-coverage additions, and regression fixes. Includes a few open-ended "Investigate X" prompts (e.g. Investigate campaign-query memory estimation gap #64583) that closed without merging.
• Example PRs: Record memory save outcomes in unified sessions and conclusion data #72133, Use Gonum helpers for grader descriptive statistics #72100, Propagate configured environments to persistence jobs #72098, Enable readerstringconv in native and wasm CI #72096, Triage setup-script security findings as false positives #71895

---

Cluster 5: Dependency & Image Version Pinning

• Size: 60 tasks (9.9% of total)
• Success Rate: 86.7% (52 merged / 8 closed / 0 open)
• Avg Comments / Reviews / Files Changed: 5.7 / 3.3 / 164.5 (far above any other cluster)
• Top Keywords: version, v0, pin, pins, image, aw, gh, gh aw
• Characteristics: Automated/semi-automated version-pin bumps across many generated files. High file count but low review friction since changes are mechanical and low-risk.
• Example PRs: Honor newer cached action pins in runtime setup #72272, Upgrade Copilot CLI to 1.0.103 with reviewed checksum pin #72099, Bump MCPG to v0.5.0 and firewall to v0.29.0 #71957, Update Pi CLI default to 1.4.0 #71899, Freeze compiler-generated Copilot SDK npm installs #71885

---

Cluster 6: PR Sous Chef Bot-Driven Fixes

• Size: 47 tasks (7.8% of total)
• Success Rate: 89.4% (42 merged / 5 closed / 0 open) — highest of any cluster
• Avg Comments / Reviews / Files Changed: 13.5 / 6.3 / 47.1 (highest engagement of any cluster)
• Top Keywords: (top TF-IDF terms for this cluster are dominated by PR Sous Chef's run-footer cost/token accounting block — "AIC", "sub" — rather than task content; inspecting raw prompts shows small, well-scoped bug/safety fixes refreshed via the bot's "Branch refresh requested by PR Sous Chef" comments)
• Characteristics: Small, targeted fixes iterated through PR Sous Chef bot comments. Highest comment/review volume and the highest merge rate — iteration converges to a mergeable state rather than stalling, consistent with prior reports.
• Example PRs: Update lethal trifecta landing page copy #65115, Prevent base config restore from clobbering an external root checkout #65069, Fail add-comment when the triggering target cannot be resolved #65068, Handle transient stdin reads in the safe-output transport #65067, Clarify PR branch target errors in scheduled runs #65066

---

Cluster 7: Work Queue / Worker System

• Size: 37 tasks (6.1% of total — smallest cluster)
• Success Rate: 83.8% (31 merged / 6 closed / 0 open)
• Avg Comments / Reviews / Files Changed: 7.7 / 5.5 / 29.5
• Top Keywords: queue, work, work queue, claim, issue, dispatch, policy, pr
• Characteristics: Work-queue/dispatch coordination logic — claim arbitration, worker dispatch, queue status reporting. Mid-pack merge rate, no red flags.
• Example PRs: Fix generated conclusion references and fetch warnings; pin Redis #71898, Fix missing work-queue diagnostic guidance #71665, [Safe Outputs] Standardize machine-readable error codes #71613, Require item-level verification in daily queue reports #71607, Report missing work queue pools with actionable recovery diagnostics #71495

Success Rate by Cluster

Cluster	Tasks	Success Rate	Open	Avg Comments	Avg Reviews	Avg Files	Top Keywords
C1: Agent/Workflow Orchestration & GitHub Integration	211	82.0%	0	5.5	3.7	46.3	agent, workflow, github, job
C2: Model Routing (Copilot/Claude/Pi)	93	86.0%	1	6.6	4.7	45.8	model, routing, copilot, claude
C3: Safe-Output Failure Handling & Diagnostics	86	83.7%	0	5.6	4.9	44.5	safe output, failure, diagnostics
C4: Ledger/Memory & Test Coverage Fixes	71	81.7%	0	4.8	2.9	13.5	ledger, memory, test, coverage
C5: Dependency & Image Version Pinning	60	86.7%	0	5.7	3.3	164.5	version, pin, image, gh aw
C6: PR Sous Chef Bot-Driven Fixes	47	89.4%	0	13.5	6.3	47.1	(bot-footer dominated; see above)
C7: Work Queue / Worker System	37	83.8%	0	7.7	5.5	29.5	queue, work, claim, dispatch

Full Data Table

All 605 analyzed PRs (click to expand)

Showing up to 15 representative PRs per cluster (605 total analyzed; full list cached at /tmp/gh-aw/cache-memory/prompt-clustering/).

PR #	Title	Cluster	Outcome
#64053	[WIP] Fix failing GitHub Actions job 'build'	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64076	[WIP] Fix failing GitHub Actions job Integration: Workflow Misc Part 2	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64089	Fix upload-assets job summary formatting	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64096	Add Daily Ecosystem Explorer agentic workflow	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64174	Add required-labels gating to assign-to-agent	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64266	Enable narrowed label schemas in triage workflows	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64304	Preserve experiment counts in usage audit reports	C1: Agent/Workflow Orchestration & GitHub Integration	Closed
#64338	Attribute friction costs to audit sources	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64339	Collect experiment and evals data in the usage artifact for audit	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64357	Filter generated files before assembling reviewer PR diffs	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64497	Restore PureLock shell utilities to PATH	C1: Agent/Workflow Orchestration & GitHub Integration	Closed
#64498	Honor runtimes.node action override in threat-detection and evals...	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64499	Document SEC-005 exemption for ledger pushes	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64805	Preserve PR review provenance for safe supersession	C1: Agent/Workflow Orchestration & GitHub Integration	Merged
#64854	Split workflow resolution into focused modules	C1: Agent/Workflow Orchestration & GitHub Integration	Closed
...	(196 more C1 PRs omitted for length)	C1: Agent/Workflow Orchestration & GitHub Integration	—
#64172	Remove implicit model defaults from fallback chains	C2: Model Routing (Copilot/Claude/Pi)	Closed
#64403	Support AWF task-level model routing in the Copilot engine	C2: Model Routing (Copilot/Claude/Pi)	Closed
#64593	Support AWF task-level model routing in the Copilot engine	C2: Model Routing (Copilot/Claude/Pi)	Merged
#64806	Add Claude 5.5 and GPT-6 pricing to model catalogs	C2: Model Routing (Copilot/Claude/Pi)	Merged
#64812	Resolve Pi auto models and classify unsupported-model errors	C2: Model Routing (Copilot/Claude/Pi)	Closed
#64859	Route model classification from workflow task text	C2: Model Routing (Copilot/Claude/Pi)	Closed
#64870	Clarify Copilot model rejection diagnostics	C2: Model Routing (Copilot/Claude/Pi)	Merged
#64969	Honor wire API overrides in Copilot SDK provider resolution	C2: Model Routing (Copilot/Claude/Pi)	Merged
#65299	Stabilize API consumption report workflow	C2: Model Routing (Copilot/Claude/Pi)	Merged
#65318	Add Codex smoke test for Copilot auto	C2: Model Routing (Copilot/Claude/Pi)	Merged
#65331	Route Codex copilot/auto to a Codex-compatible model	C2: Model Routing (Copilot/Claude/Pi)	Closed
#65344	Normalize unified agent session metadata and AIC accounting	C2: Model Routing (Copilot/Claude/Pi)	Merged
#65404	Resolve Codex model rejections in eight workflows	C2: Model Routing (Copilot/Claude/Pi)	Closed
#65877	Route Pi models to compatible inference APIs	C2: Model Routing (Copilot/Claude/Pi)	Merged
#66234	Separate routing candidates from sub-agent model policy	C2: Model Routing (Copilot/Claude/Pi)	Merged
...	(78 more C2 PRs omitted for length)	C2: Model Routing (Copilot/Claude/Pi)	—
#64176	Preserve resolvable safe-output targets	C3: Safe-Output Failure Handling & Diagnostics	Merged
#64356	Honor add-labels.max above 10 for per-call label limit	C3: Safe-Output Failure Handling & Diagnostics	Closed
#64496	Disable credential persistence for safe-output checkouts	C3: Safe-Output Failure Handling & Diagnostics	Closed
#65065	Clarify explicit PR targets rejected by triggering-only push config...	C3: Safe-Output Failure Handling & Diagnostics	Closed
#65126	Refactor coordinator smoke verification to JavaScript	C3: Safe-Output Failure Handling & Diagnostics	Closed
#65302	Standardize session handler error codes	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65402	Route Windows CLI integration failures through safe outputs	C3: Safe-Output Failure Handling & Diagnostics	Closed
#65406	Surface zero-token driver exits and repair Avenger and PR Triage st...	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65408	Skip failure issues for invalidated PR merge-ref checkouts	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65409	Fix AOAI smoke issue targeting and clarify PR safe-output limits	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65501	Avoid false safe-output rate-limit warnings	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65561	Fix issue-clustering dashboard failure and clipped issue text	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65562	Harden Pi provider retries and failure diagnostics	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65564	Emit incomplete outcomes with diagnostics when agents produce no sa...	C3: Safe-Output Failure Handling & Diagnostics	Merged
#65571	Remove dead close_discussion instructions from daily reports	C3: Safe-Output Failure Handling & Diagnostics	Merged
...	(71 more C3 PRs omitted for length)	C3: Safe-Output Failure Handling & Diagnostics	—
#64199	[WIP] Add distributed ledger capability to repo-memory	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64337	Move daily workflow review history to repo-memory ledgers	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64473	Allow arXiv researcher to update its ledger and dedup cache	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64501	Recognize consequence-based empty-catch rationale	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64502	Remove dynamic eval from update-release tests	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64581	Add declarative ledger types for collection replay and latest-wins ...	C4: Ledger/Memory & Test Coverage Fixes	Closed
#64583	Investigate campaign-query memory estimation gap	C4: Ledger/Memory & Test Coverage Fixes	Closed
#64599	Move ledger compaction into Agentic Maintenance (untrusted plan job...	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64709	Consolidate ledger compaction into two Agentic Maintenance jobs	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64721	Add built-in log, set, map, table, and counter ledgers	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64813	Expose operation-specific tools for map and work-pool ledgers	C4: Ledger/Memory & Test Coverage Fixes	Merged
#64850	Remove work-pool ledger support	C4: Ledger/Memory & Test Coverage Fixes	Merged
#65055	Document versioned dispatch coordinator ledger upgrades	C4: Ledger/Memory & Test Coverage Fixes	Merged
#65387	Reduce test fanout on pull requests	C4: Ledger/Memory & Test Coverage Fixes	Merged
#65405	Restore ledger push handoff and correct smoke projection checks	C4: Ledger/Memory & Test Coverage Fixes	Merged
...	(56 more C4 PRs omitted for length)	C4: Ledger/Memory & Test Coverage Fixes	—
#64016	Bump gh-aw-mcpg to v0.4.27 and gh-aw-firewall to v0.28.27	C5: Dependency & Image Version Pinning	Merged
#64127	Update generated Claude Code CLI to 2.1.280	C5: Dependency & Image Version Pinning	Merged
#64268	Manage self-hosted runner enforcement with gh aw env	C5: Dependency & Image Version Pinning	Closed
#64336	Write version output to stdout	C5: Dependency & Image Version Pinning	Merged
#64359	Update agentic engine CLI pins and recompile workflows	C5: Dependency & Image Version Pinning	Merged
#64500	Update stale CLI and scanner image pins	C5: Dependency & Image Version Pinning	Merged
#64871	Add GPT-6.1 Sol support to Copilot workflows	C5: Dependency & Image Version Pinning	Merged
#64872	Refresh default agent CLI and Copilot SDK versions	C5: Dependency & Image Version Pinning	Merged
#64917	Bump gh-aw-firewall and gh-aw-mcpg versions	C5: Dependency & Image Version Pinning	Merged
#65095	Use the gh.io/gh-aw short URL in docs homepage social cards	C5: Dependency & Image Version Pinning	Merged
#65106	Version and upgrade dispatch coordinator ledger messages	C5: Dependency & Image Version Pinning	Merged
#65614	Harden Windows CLI integration workflow	C5: Dependency & Image Version Pinning	Closed
#65661	Consolidate container image CVE tracking	C5: Dependency & Image Version Pinning	Merged
#65929	Harden Windows CLI integration workflow	C5: Dependency & Image Version Pinning	Merged
#66121	Harden installer downloads and security ownership paths	C5: Dependency & Image Version Pinning	Closed
...	(45 more C5 PRs omitted for length)	C5: Dependency & Image Version Pinning	—
#64023	Add five trajectory graders and harden generator contracts	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64031	Allow two noop calls by default	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64052	Render friction cost in workflow step summaries	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64093	Repoint pydantic-ai engine import from archived pydantic-ai-harness...	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64173	Enforce self-hosted runners during compilation	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64175	Preserve version labels on SHA-pinned workflow actions	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64177	Infer Copilot wire API for utility model variants	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64179	Allow workflows to narrow add-labels item schemas	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64180	Automatically allow OTLP collector endpoints through the firewall	C6: PR Sous Chef Bot-Driven Fixes	Closed
#64267	Harden private-to-public flow validation	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64302	Normalize sandbox frontmatter fields to kebab-case	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64334	Use GitHub App token for gh-proxy	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64335	Handle malformed aw_context in manual dispatches	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64354	Add standalone safe-output-backed ledger configuration	C6: PR Sous Chef Bot-Driven Fixes	Merged
#64420	Add replay projections to standalone ledgers	C6: PR Sous Chef Bot-Driven Fixes	Merged
...	(32 more C6 PRs omitted for length)	C6: PR Sous Chef Bot-Driven Fixes	—
#64853	Dispatch Work Coordinator — implementation pending	C7: Work Queue / Worker System	Closed
#65062	Add Git-backed work queue operator commands	C7: Work Queue / Worker System	Merged
#65097	Rename work coordinator tool and CLI command to work-queue	C7: Work Queue / Worker System	Merged
#65103	Instrument dispatch coordinator queue and preserve smoke failure re...	C7: Work Queue / Worker System	Merged
#65192	Include WorkQ claims in the admission prompt	C7: Work Queue / Worker System	Merged
#65384	Document work queue orchestration and inspection	C7: Work Queue / Worker System	Merged
#65386	Route ESLint factory tasks through an Issues-backed queue	C7: Work Queue / Worker System	Closed
#65388	Add daily work queue status report	C7: Work Queue / Worker System	Merged
#65443	Dispatch queued work to trusted worker workflows	C7: Work Queue / Worker System	Merged
#65492	Keep the essential-issue dashboard current when discussion edits ar...	C7: Work Queue / Worker System	Merged
#65494	Add selectable issue-backed work queue storage	C7: Work Queue / Worker System	Merged
#65495	Add declarative sorting to work queue recommendations	C7: Work Queue / Worker System	Merged
#65591	Migrate ESLint factory workflows to work-queue dispatch	C7: Work Queue / Worker System	Merged
#66950	Prevent blank space in documentation tables	C7: Work Queue / Worker System	Merged
#66991	Scope work queue stress runs and shorten PR profiles	C7: Work Queue / Worker System	Merged
...	(22 more C7 PRs omitted for length)	C7: Work Queue / Worker System	—

Key Findings

1. No outlier cluster this cycle — a reversal from the 2026-10-26 report. That report flagged a 63.5%-merge "gh-aw Compiler/CLI Core & Infra Hardening" cluster (52 PRs, 16.2 points below its 79.7% overall rate), driven mostly by silently-abandoned "Investigate/Assess/Document"-style prompts. This run's merge rates are tightly clustered (81.7%–89.4%, all within ~8 points of the 84.0% overall rate). Related firewall/infra-hardening PRs are now folded into the much larger C1 cluster, which sits at 82.0% — near the floor of this run's range but not an outlier by the sizing/gap thresholds.
2. High engagement still correlates with higher merge rates, not lower. C6 (PR Sous Chef Bot-Driven Fixes) has the highest comment/review volume (13.5/6.3 avg) and the highest merge rate (89.4%) — the same pattern seen in prior reports: bot-driven iteration converges rather than stalls.
3. Bulk mechanical changes remain low-risk despite size. C5 (Dependency & Image Version Pinning) touches ~165 files per PR on average — 3-12x any other cluster — yet merges at 86.7%, the second-highest rate, confirming diff size isn't itself a risk signal when the change is mechanical and narrowly scoped.
4. The largest cluster converges reliably. C1 (Agent/Workflow Orchestration & GitHub Integration) is more than double the size of the next-largest cluster (211 vs. 93 PRs) yet still merges at 82.0%, in line with the rest of the distribution — broad topic diversity isn't suppressing success rate.
5. Virtually everything in the window has resolved. Only 1 of 605 PRs remains open (Support Claude on Vertex AI with GCP WIF #72273, Model Routing), vs. 37 open in the 2026-10-26 report's 592-PR window — expected given the window has rotated ~2 days forward and previously-open PRs had more time to resolve; not a like-for-like improvement signal (see Limitations).

Lowest-Merge Cluster Root Cause

Not applicable this run. The lowest-performing cluster (C4: Ledger/Memory & Test Coverage Fixes, 81.7% merged) is only 2.3 points below the 84.0% overall rate — well short of the 10-point gap threshold that triggers a root-cause drill-down, regardless of its size (71 PRs, above the 15-PR size floor). No cluster in this run meets both outlier criteria (size ≥ 15 and gap ≥ 10 points), so Phase 4.5's blocker-classification step was not run. This methodology will automatically re-engage the next time a qualifying outlier appears.

Recommendations

1. No remediation needed this cycle. Every cluster is converging at a healthy rate (82–89%); current prompt-scoping and bot-iteration practices (PR Sous Chef refresh comments) appear to be working across the board.
2. Keep leaning on PR Sous Chef-style iterative refinement. C6's pattern (highest engagement + highest merge rate) is the strongest repo-wide signal for "what makes a task recoverable" — consider extending similar automated follow-up nudges to clusters with lower review/comment volume (e.g. C4, C1) if their merge rates drift down in future runs.
3. Watch C1 even though it isn't flagged. It's both the largest cluster (35% of all tasks) and sits at the bottom of this cycle's merge-rate range (82.0%) — the same thematic territory (firewall/infra/compiler-adjacent glue work) that produced last cycle's outlier. Re-running this analysis regularly will catch it early if it drifts toward outlier status again.
4. Avoid open-ended "Investigate/Assess/Document" prompts. C4 still contains a few closed PRs from this pattern (e.g. Investigate campaign-query memory estimation gap #64583, "Investigate campaign-query memory estimation gap"); converting these into concrete, scoped implementation asks remains good practice even though the cluster isn't an outlier this time.

Limitations

• No workflow-run turn/duration/cost metrics are included this run. gh-aw's own workflow-run logs (via the agenticworkflows logs tool) track runs of this repository's own agentic workflows under the copilot engine, which do not correspond 1:1 to the Copilot coding-agent PRs being clustered here — this was established as a dead end in the 2026-10-26 run and was not re-attempted this cycle.
• Clustering silhouette score (0.054) is low in absolute terms, reflecting that these prompts share substantial vocabulary across topics (all are gh-aw engineering tasks) rather than being cleanly separable domains — cluster boundaries should be read as thematic groupings, not hard categories.
• The 79.7% → 84.0% overall merge-rate change vs. the 2026-10-26 report is not a verified trend: the 30-day window rotated forward by ~2 days (dropping the oldest ~13 PRs and adding newly-created ones), and PRs that were still open in the prior snapshot had two more days to resolve by this run. A like-for-like comparison would require re-scoring the exact same PR set at both points in time.
• Full data table capped at 15 PRs per cluster for readability; the complete 605-PR list with cluster assignments is cached at /tmp/gh-aw/cache-memory/prompt-clustering/full-pr-table-2026-10-28.json for future runs.

---

Generated by Prompt Clustering Analysis (Run: §40666371572)

References:

• §40666371572 — this analysis run

Generated by 📊 Copilot Agent Prompt Clustering Analysis · claude · agent · ~163 AIC · ⌖ ~1 AIC · ⊞ 5K · ◷

• expires on Oct 29, 2026, 2:08 AM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Copilot Agent Prompt Clustering Analysis.

A newer discussion is available at Discussion #72919.