[prompt-clustering] Copilot Agent Prompt Clustering Analysis — 2026-02-21

[github-actions[bot]]

Daily NLP-based clustering analysis of Copilot coding agent task prompts from the last 30 days.

Summary

Metric	Value
Analysis Period	Last 30 days
Total PRs Analyzed	896
Clusters Identified	10
Overall Merge Rate	68.9%
Peak Week	2026-W05 (394 PRs)
Avg PRs/Week	299
Most Common Task Type	issue_resolution (361)

Outcome distribution: merged: 617, closed: 275, open: 4

Cluster Overview

Cluster	Theme	PRs	Merge Rate	Top Keywords
C2	Workflow & CI/CD	170	59%	issue, section, workflow, gh
C4	Dependency & Package Updates	166	78%	update, add, agent, github
C10	Issue Resolution	108	60%	quality, code, code quality, files
C1	Bug Fix & Testing	98	72%	url, reference, debug, fix
C3	Workflow & CI/CD	89	79%	agentic, agentic workflows, workflows, workflow
C5	Dependency & Package Updates	83	63%	mcp, server, mcp server, gateway
C8	Safe Outputs / Project Tasks	72	72%	project, safe, safe outputs, outputs
C7	Campaign & Security	52	67%	campaign, security, project, url
C9	CI Failure Fixes	31	81%	job, fix, id, implement
C6	Agentic Failure Debugging	27	59%	failure, debug, agent, workflow

Key Findings

1. Workflow & CI/CD dominates (170 PRs, 19% of total): The most common task category centers on issue, section, workflow. This suggests a high volume of routine issue-driven maintenance work delivered via structured issue resolution prompts.

2. CI Failure Fixes have the highest success rate (81% merge rate): Prompts asking the agent to fix a specific failing CI job with a job ID and URL tend to produce code that reviewers accept with minimal friction — these are well-scoped, bounded tasks.

3. MCP Server tasks involve the most files (avg 37.1 files changed): MCP dependency updates touch broad package manifests and lock files, requiring more careful review and coordination.

4. Task type distribution: issue_resolution tasks are the most common at 40%, followed by workflow_change at 14% and bug_fix at 13%.

5. Two distinct issue-resolution patterns: Cluster C2 (structured (issue_description) blocks, 59% merge rate) vs Cluster C10 (code quality/refactoring issues, 60% merge rate) — both show that generic issue dumps yield lower success than tightly scoped tasks.

Recommendations

Based on clustering analysis:

1. Standardize high-volume task prompts: The Workflow & CI/CD cluster has 170 PRs — consider creating prompt templates for these common patterns to improve consistency and reduce agent iterations.

2. Leverage CI failure prompt structure: The CI Failure Fixes cluster (81% merge rate) uses a highly structured prompt with job ID, job URL, and explicit instructions — this pattern should be adopted more broadly.

3. Decompose complex MCP update tasks: The MCP Server cluster (avg 37.1 files changed, 63% merge rate) may benefit from breaking tasks into smaller, more focused units to improve reviewability.

4. Capture issue context upfront: Many prompts in Clusters C2 and C6 reference GitHub issues by URL only — enriching prompts with issue body text in the prompt itself could reduce clarification iterations and improve the 59% merge rates.

Detailed Cluster Analysis

Cluster 2: Workflow & CI/CD (Issue-Driven)

Metric	Value
Size	170 PRs (19% of total)
Merge Rate	59%
Avg Files Changed	10.5
Avg Code Additions	312
Avg Comments	1.9

Top Keywords: issue · section · workflow · gh · gh aw · aw

Task Types: issue_resolution: 168, dependency_update: 1, feature_add: 1

Example PRs: [#12085], [#12130], [#12574]

Sample prompts from this cluster

---- This section details on the original issue you should resolve (issue_title)[CI Failure Doctor] Test failures after PR #11036: regex pattern changes without test updates...
---- This section details on the original issue you should resolve (issue_title)[deep-report] Install Go toolchain in Daily CLI Performance Agent workflow...
---- This section details on the original issue you should resolve (issue_title) Agentic Maintenance improvements — merge close issues and close discussions in same batch...

---

Cluster 4: Dependency & Package Updates (General)

Metric	Value
Size	166 PRs (19% of total)
Merge Rate	78%
Avg Files Changed	27.3
Avg Code Additions	362
Avg Comments	4.2

Top Keywords: update · add · agent · github · use · remove

Task Types: documentation: 34, other: 33, workflow_change: 32

Example PRs: [#12628], [#12062], [#12093]

Sample prompts from this cluster

Update the init command behavior as follows: When the command is invoked without arguments, it should enter an interactive mode and prompt the user to select which agent engine to use: Copilot, Cla...
There should be no more mention or distinction between active and passive campaigns anymore. They are all active in that they orchestrator makes decisions and dispatches.
format, lint js, fix js tests

---

Cluster 10: Issue Resolution (Code Quality)

Metric	Value
Size	108 PRs (12% of total)
Merge Rate	60%
Avg Files Changed	11.8
Avg Code Additions	577
Avg Comments	0.9

Top Keywords: quality · code · code quality · files · discussion · issue

Task Types: issue_resolution: 108

Example PRs: [#13182], [#12076], [#13005]

Sample prompts from this cluster

---- [testify-expert] Improve Test Quality: pkg/workflow/strings_test.go — The test file lacks table-driven tests and comprehensive assertions...
---- [ci-coach] Optimize integration test matrix by isolating slow Docker build test...
---- [refactor] Semantic Function Clustering Analysis - Refactoring Opportunities in pkg/workflow...

---

Cluster 1: Bug Fix & Testing (URL/Reference Driven)

Metric	Value
Size	98 PRs (11% of total)
Merge Rate	72%
Avg Files Changed	20.6
Avg Code Additions	900
Avg Comments	2.2

Top Keywords: url · reference · debug · fix · review · run

Task Types: bug_fix: 39, documentation: 20, other: 13

Example PRs: [#12864], [#11567], [#13063]

Sample prompts from this cluster

Investigate why the javascript code region balancer function was not able to balance the description in this issue: #11081
Review docs mentioning "assign to copilot" and make sure they match the implementation of the code and (githubnext.github.io/redacted)
Reference: https://github.com/githubnext/gh-aw/actions/runs/21230742993/job/... review the reference workflow-run error and update server.sh to use the new file location.

---

Cluster 3: Workflow & CI/CD (Agentic Workflows)

Metric	Value
Size	89 PRs (10% of total)
Merge Rate	79%
Avg Files Changed	16.8
Avg Code Additions	404
Avg Comments	2.9

Top Keywords: agentic · agentic workflows · workflows · workflow · md · create

Task Types: workflow_change: 56, documentation: 15, other: 8

Example PRs: [#12589], [#12453], [#11360]

Sample prompts from this cluster

Update the template used to create the parent issue for all agentic-workflow issues so that it creates a conclusion job. Set the issue title to "agentic-workflows failures"...
If you update the code that checks whether workflows are in sync in the agentic maintenance workflow, and the gh-aw-agent-token secret is available, then the workflow should...
Update issue monster agentic workflow to ignore issues that already have assignees.

---

Cluster 5: MCP Server & Gateway

Metric	Value
Size	83 PRs (9% of total)
Merge Rate	63%
Avg Files Changed	37.1
Avg Code Additions	499
Avg Comments	4.7

Top Keywords: mcp · server · mcp server · gateway · mcp gateway · tool

Task Types: issue_resolution: 45, workflow_change: 8, bug_fix: 8

Example PRs: [#11929], [#11632], [#11928]

Sample prompts from this cluster

Run the update command and ensure that the Sentry MCP is updated. It should be upgraded to version 0.27.0, as shown in the dependency pull-request description attached below.
---- [deep-report] Fix GitHub Remote MCP Authentication Test tool availability...
---- [ca] CLI Version Updates: Copilot 0.0.388, Sandbox Runtime 0.0.29, MCP Gateway v0.0.74...

---

Cluster 8: Safe Outputs / Project Tasks

Metric	Value
Size	72 PRs (8% of total)
Merge Rate	72%
Avg Files Changed	19.4
Avg Code Additions	411
Avg Comments	4.4

Top Keywords: project · safe · safe outputs · outputs · update · create

Task Types: bug_fix: 27, feature_add: 16, documentation: 6

Example PRs: [#11120], [#12941], [#12690]

Sample prompts from this cluster

---- [safe-output] create-pull-request is failing — since we moved safe-outputs in its own container...
Update safe-outputs mcp configuration to use node:lts-slime and mount the git bin folder to make git accessible to the container.
Create a Dockerfile for safe-outputs that installs git in the container (node alpine) and start the server.

---

Cluster 7: Campaign & Security

Metric	Value
Size	52 PRs (6% of total)
Merge Rate	67%
Avg Files Changed	9.5
Avg Code Additions	437
Avg Comments	2.0

Top Keywords: campaign · security · project · url · issue · new

Task Types: bug_fix: 12, workflow_change: 10, documentation: 9

Example PRs: [#11178], [#11232], [#12053]

Sample prompts from this cluster

Don't rely on cache memory for campaign discovery but use labels. Each campaign issue (epic or worker) should get a label "agentic-campaign"...
Don't require a tracker-id for campaign worker workflows...
For campaigns, make workers first-class "campaign workers" and keep orchestration concerns explicit, rather than relying on fusion as a permanent crutch...

---

Cluster 9: CI Failure Fixes (Job-ID Driven)

Metric	Value
Size	31 PRs (3% of total)
Merge Rate	81%
Avg Files Changed	28.9
Avg Code Additions	2536
Avg Comments	0.4

Top Keywords: job · fix · id · implement · failing · url

Task Types: test_fix: 15, bug_fix: 13, workflow_change: 1

Example PRs: [#13361], [#13949], [#13353]

Sample prompts from this cluster

Fix the failing GitHub Actions workflow js — Analyze the workflow logs, identify the root cause of the failure, and implement a fix. Job ID: 61070763482 Job URL: ...
Fix the failing GitHub Actions workflow lint-go — Analyze the workflow logs, identify the root cause of the failure, and implement a fix. Job ID: 61501187403 Job URL: ...

---

Cluster 6: Agentic Failure Debugging

Metric	Value
Size	27 PRs (3% of total)
Merge Rate	59%
Avg Files Changed	9.9
Avg Code Additions	200
Avg Comments	2.1

Top Keywords: failure · debug · agent · workflow · failed · action

Task Types: issue_resolution: 27

Example PRs: [#13855], [#14044], [#11093]

Sample prompts from this cluster

---- [agentics] Instructions Janitor failed — Workflow: [Instructions Janitor]...
---- [agentics] Code Scanning Fixer failed — Workflow: [Code Scanning Fixer]...
---- [agentics] Agentic Campaign Generator failed — Workflow: [Agentic Campaign Generator]...

Task Type Distribution

Task Type	Count	% of Total
issue_resolution	361	40.3%
workflow_change	121	13.5%
bug_fix	118	13.2%
documentation	88	9.8%
feature_add	72	8.0%
other	69	7.7%
test_fix	33	3.7%
dependency_update	26	2.9%
refactor	8	0.9%

Recent PRs Sample (Last 50)

PR #	Title	Cluster	Outcome	Top Keywords
#14275	[WIP] Improve test quality for threat detection fi...	C10: Issue Resolution	❌ open	quality, code, code quality
#14274	[WIP] Fix firewall SSL-bump configuration extracti...	C10: Issue Resolution	❌ open	quality, code, code quality
#14273	[WIP] Add documentation for health command in CLI	C10: Issue Resolution	❌ open	quality, code, code quality
#14269	Revert gh-aw-mcpg to v0.0.103	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14268	Add workflow guidance and cross-references to CLI...	C10: Issue Resolution	❌ closed	quality, code, code quality
#14267	Verify @playwright/mcp version is already updated...	C5: MCP Server & Gateway	❌ closed	mcp, server, mcp server
#14266	Document SSL-bump feature for AWF firewall	C10: Issue Resolution	✅ merged	quality, code, code quality
#14260	Add fuzzy matching "did you mean" suggestions for...	C10: Issue Resolution	✅ merged	quality, code, code quality
#14259	Investigation: CI failure #14239 is false alarm...	C2: Workflow & CI/CD	❌ closed	issue, section, workflow
#14257	Fix compiler obfuscation: Don't wrap static quoted...	C10: Issue Resolution	✅ merged	quality, code, code quality
#14255	Fix issue-monster workflow by enabling needs.* exp...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14253	Document Actions permission restrictions detected...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14225	Refactor Dependabot Project Manager to process PRs...	C8: Safe Outputs / Project	✅ merged	project, safe, safe outputs
#14222	Remove anonymous bash tool syntax, require explici...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14221	Fix TestRuntimeSetupPreservesUserVersions false po...	C9: CI Failure Fixes	✅ merged	job, fix, id
#14211	Add edit tool and full bash access to daily-cli-to...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14209	Make upgrade command version check non-blocking...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14208	Wrap agent log rendering in collapsible details...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14201	Add network access to agentic-workflows MCP server...	C5: MCP Server & Gateway	✅ merged	mcp, server, mcp server
#14191	Format auto-added "Fixes #N" as bullet point in PR...	C4: Dependency & Package Updates	✅ merged	update, add, agent
#14189	Teach create workflow agent to discover CLI automa...	C8: Safe Outputs / Project	✅ merged	project, safe, safe outputs
#14184	Limit create-issue to 1 in daily-cli-tools-tester...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14182	Add Dependabot Project Manager workflow with bundl...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14173	Rename MCP server identifier from agentic_workflow...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14171	Fix smoke-claude: handle tool failures gracefully...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14168	Add daily exploratory testing workflow for CLI too...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14156	Capture exit codes and stderr when gh CLI commands...	C5: MCP Server & Gateway	✅ merged	mcp, server, mcp server
#14150	Remove prompt file management functions from init/...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14149	Add max-tokens and max-iterations execution bounds...	C4: Dependency & Package Updates	❌ closed	update, add, agent
#14147	Add daily concurrency analysis workflow for MCP se...	C3: Agentic Workflows	✅ merged	agentic, agentic workflows, workflows
#14140	Add binary path detection for MCP server self-invo...	C5: MCP Server & Gateway	✅ merged	mcp, server, mcp server
#14129	Add unit tests for compiler_yaml_main_job.go	C10: Issue Resolution	✅ merged	quality, code, code quality
#14127	Fix daily-fact workflow action-tag to include miss...	C5: MCP Server & Gateway	✅ merged	mcp, server, mcp server
#14125	Fix add_comment to handle discussion numbers via f...	C1: Bug Fix & Testing	✅ merged	url, reference, debug
#14121	Investigation: Daily Issues Report Generator fail...	C6: Agentic Failure Debugging	❌ closed	failure, debug, agent
#14120	Fix gh-aw binary availability for user-defined ste...	C6: Agentic Failure Debugging	✅ merged	failure, debug, agent
#14119	Fix textblob version check in Copilot PR NLP Analy...	C6: Agentic Failure Debugging	✅ merged	failure, debug, agent
#14114	Update CLI versions: Claude Code 2.1.34, Copilot 0...	C5: MCP Server & Gateway	✅ merged	mcp, server, mcp server
#14095	Replace emojis with Starlight icons in documentati...	C1: Bug Fix & Testing	❌ closed	url, reference, debug
#14098	Consolidate monitoring pattern into ProjectOps	C4: Dependency & Package Updates	❌ closed	update, add, agent

References: §22256749758

AI generated by Copilot Agent Prompt Clustering Analysis

• expires on Feb 28, 2026, 12:43 PM UTC

[github-actions[bot]]

💥 WHOOSH!

The smoke test agent swoops in from the shadows of the CI pipeline...

⚡ KAZAM! The Claude engine has been here — and all systems are GO!

🦸 SMOKE TEST AGENT — Run §22256990263 — was here!

POW! BAM! ZAPP! All core checks passed. The agentic workflows live to fight another day. 🎉

...and just like that, the agent vanishes back into the GitHub Actions cloud.

💥 [THE END] — Illustrated by Smoke Claude

[github-actions[bot]]

🤖 Beep boop! The smoke test agent was here! 👋

I just swung by to verify that the copilot engine is running smoother than a freshly compiled binary. Tests are passing, neurons are firing, and automation is doing its beautiful thing!

✨ Stay compilable, my friends. ✨

📰 BREAKING: Report filed by Smoke Copilot