[prompt-clustering] Daily Copilot Agent Prompt Clustering Analysis - 2026-02-10

[github-actions[bot]]

🔬 Copilot Agent Prompt Clustering Analysis

Analysis Date: 2026-02-10
Analysis Period: Last 30 days
Total Tasks Analyzed: 1,182
Clusters Identified: 10

Summary

• Total PRs Analyzed: 1,182
• Merged: 781 (66.1%)
• Closed (not merged): 391
• Open: 11
• Clusters: 10

View Full Analysis

Cluster Analysis

Cluster 1: Dependency Updates & Version Bumps (23.6% of tasks)

• Size: 279 tasks
• Success Rate: 73.8% (206 merged)
• Avg Files Changed: 19.7
• Avg Additions: 412.1 lines
• Avg Deletions: 256.6 lines
• Top Keywords: workflow, add, update, test, docs
• Top Terms: workflow, cli, code, campaign, command

Example PRs:

• Implement deterministic SHA-256 hash for workflow frontmatter with imports, template expressions, versions, and hash-based stale detection #12589: Implement deterministic SHA-256 hash for workflow frontmatter with imports
• Add plugin installation support via frontmatter with dual-format configuration and cascading token resolution #14041: Add plugin installation support via frontmatter with dual-format configuration
• Parallelize setup operations with engine-agnostic installation #12574: Parallelize setup operations with engine-agnostic installation

Sample Task Titles

• Install Go toolchain in daily-cli-performance workflow
• chore: campaign discovery via label-based approach
• Disallow invalid permissions: read shorthand, require read-all
• Fix markdown code region balancer: distinguish consecutive blocks from nesting
• Remove invalid campaign-manager.md reference from developer instructions

---

Cluster 2: Dependency Updates & Version Bumps (10.6% of tasks)

• Size: 125 tasks
• Success Rate: 69.6% (87 merged)
• Avg Files Changed: 23.2
• Avg Additions: 543.4 lines
• Avg Deletions: 469.4 lines
• Top Keywords: workflow, add, update, test, bug
• Top Terms: workflows, github, agentic, workflow, md

Example PRs:

• Add global mood control file for centralized agent behavior #14380: Add global mood control file for centralized agent behavior
• Merge main and recompile workflows after test-yaml-import relocation #13780: Merge main and recompile workflows after test-yaml-import relocation
• Add build step and timeout updates to smoke workflows #11968: Add build step and timeout updates to smoke workflows

---

Cluster 3: MCP Server & Gateway Updates (10.5% of tasks)

• Size: 124 tasks
• Success Rate: 62.9% (78 merged)
• Avg Files Changed: 31.0
• Avg Additions: 339.3 lines
• Avg Deletions: 731.5 lines
• Top Keywords: workflow, add, test, config, update
• Top Terms: mcp, server, gateway, mcp server, tool

Example PRs:

• Add payloadDir field to MCP gateway spec for shared large payload access #13167: Add payloadDir field to MCP gateway spec for shared large payload access
• Update CLI/MCP tool versions: Copilot 0.0.394, Claude 2.1.19, Codex 0.89.0, Playwright v1.58.0, MCP Gateway v0.0.78 #11632: Update CLI/MCP tool versions: Copilot 0.0.394, Claude 2.1.19, Codex 0.89.0
• Update CLI tool versions: Claude 2.1.31, Copilot 0.0.402, GitHub MCP v0.30.3, Playwright 0.0.63, MCP Gateway v0.0.98 #13645: Update CLI tool versions: Claude 2.1.31, Copilot 0.0.402, GitHub MCP v0.30.3

---

Cluster 4: Bug Fixes & Error Resolution (9.6% of tasks)

• Size: 114 tasks
• Success Rate: 74.6% (85 merged) ✅ Best Success Rate
• Avg Files Changed: 13.8
• Avg Additions: 929.1 lines
• Avg Deletions: 175.5 lines
• Top Keywords: test, workflow, bug, add, fix
• Top Terms: test, error, tests, workflow, compiler

Example PRs:

• Refactor ParseWorkflowFile: 383→74 lines + remove network.mode concept #11671: Refactor ParseWorkflowFile: 383→74 lines + remove network.mode concept
• Update expected hash in frontmatter cross-language validation test #13353: Update expected hash in frontmatter cross-language validation test
• [WIP] Fix failing GitHub Actions workflow CI / test (pull_request) #12647: [WIP] Fix failing GitHub Actions workflow CI / test (pull_request)

---

Cluster 5: Agent & Workflow Infrastructure (9.4% of tasks)

• Size: 111 tasks
• Success Rate: 43.2% ⚠️ Lowest Success Rate
• Avg Files Changed: 11.8
• Avg Additions: 207.2 lines
• Avg Deletions: 506.2 lines
• Top Keywords: add, update, workflow, docs, test
• Top Terms: agent, details, summary, coding agent, coding

Example PRs:

• Use runtime-import for markdown body instead of inlining (with recursive processing) #12687: Use runtime-import for markdown body instead of inlining (with recursive processing)
• Update AWF to v0.13.0 and enable --enable-chroot #12827: Update AWF to v0.13.0 and enable --enable-chroot
• Update AWF firewall to v0.11.2 with ACT agent container #11567: Update AWF firewall to v0.11.2 with ACT agent container

Note: High proportion of [WIP] PRs in this cluster, indicating complexity and experimentation.

---

Cluster 6: Issue & Workflow Management (9.4% of tasks)

• Size: 111 tasks
• Success Rate: 51.4%
• Avg Files Changed: 11.8
• Avg Additions: 238.5 lines
• Avg Deletions: 45.7 lines
• Top Keywords: bug, workflow, add, fix, docs
• Top Terms: issue, workflow, issues, title, details

Example PRs:

• Use AWF --enable-chroot mode and remove unnecessary --mount and --env flags #13576: Use AWF --enable-chroot mode and remove unnecessary --mount and --env flags
• Update AWF (gh-aw-firewall) to v0.13.1 #13378: Update AWF (gh-aw-firewall) to v0.13.1
• Emit workflow-level permissions from frontmatter instead of always using empty block #14461: Emit workflow-level permissions from frontmatter instead of always using empty block

---

Cluster 7: Safe Outputs & API Additions (8.5% of tasks)

• Size: 101 tasks
• Success Rate: 74.3%
• Avg Files Changed: 26.9
• Avg Additions: 511.6 lines
• Avg Deletions: 185.8 lines
• Top Keywords: add, workflow, bug, api, test
• Top Terms: safe, output, outputs, safe outputs, agent

Example PRs:

• Propagate create-discussion errors to conclusion job failure tracking #13399: Propagate create-discussion errors to conclusion job failure tracking
• Add git credentials cleanup and regeneration for agent execution #14700: Add git credentials cleanup and regeneration for agent execution
• Remove verbose "All pinned actions are up to date" message from stderr #13350: Remove verbose "All pinned actions are up to date" message from stderr

---

Cluster 8: CLI Command & Tooling (7.0% of tasks)

• Size: 83 tasks
• Success Rate: 65.1%
• Avg Files Changed: 22.2
• Avg Additions: 1,058.5 lines
• Avg Deletions: 988.2 lines
• Top Keywords: workflow, update, test, add, docs
• Top Terms: gh, aw, gh aw, command, workflow

Example PRs:

• [WIP] Update to gh-aw-mcpg version 0.0.113 #14559: [WIP] Update to gh-aw-mcpg version 0.0.113
• Update MCP gateway to v0.0.94 #13360: Update MCP gateway to v0.0.94
• Revert MCP Gateway to v0.0.78 #14264: Revert MCP Gateway to v0.0.78

---

Cluster 9: Project Management & Campaigns (6.0% of tasks)

• Size: 71 tasks
• Success Rate: 64.8%
• Avg Files Changed: 8.4
• Avg Additions: 408.4 lines
• Avg Deletions: 294.6 lines
• Top Keywords: add, update, workflow, api, bug
• Top Terms: project, update, campaign, create, safe

Example PRs:

• Upgrade actions/download-artifact to v7.0.0 to eliminate Buffer() deprecation warnings #13154: Upgrade actions/download-artifact to v7.0.0 to eliminate Buffer() deprecation warnings
• [WIP] Split update and create project status update #12961: [WIP] Split update and create project status update
• [WIP] Remove update-project and create-project-update-status outputs #12941: [WIP] Remove update-project and create-project-update-status outputs

---

Cluster 10: Documentation & Code Quality (5.3% of tasks)

• Size: 63 tasks
• Success Rate: 71.4%
• Avg Files Changed: 5.7
• Avg Additions: 289.6 lines
• Avg Deletions: 53.1 lines
• Top Keywords: docs, add, workflow, bug, update
• Top Terms: documentation, files, workflows, code, campaign

Example PRs:

• Fix package documentation by removing duplicate package comments from 70 files #12301: Fix package documentation by removing duplicate package comments from 70 files
• Bundle npm dependency updates for docs #12641: Bundle npm dependency updates for docs
• Apply strict matching to slash commands (startsWith + exact equality) #14702: Apply strict matching to slash commands (startsWith + exact equality)

Success Rate by Cluster

Cluster	Description	Tasks	Success Rate	Avg Files	Top Keywords
4	Bug Fixes & Error Resolution	114	74.6% ✅	13.8	test, workflow, bug
7	Safe Outputs & API Additions	101	74.3%	26.9	add, workflow, bug
1	Dependency Updates & Version Bumps	279	73.8%	19.7	workflow, add, update
10	Documentation & Code Quality	63	71.4%	5.7	docs, add, workflow
2	Dependency Updates & Version Bumps	125	69.6%	23.2	workflow, add, update
8	CLI Command & Tooling	83	65.1%	22.2	workflow, update, test
9	Project Management & Campaigns	71	64.8%	8.4	add, update, workflow
3	MCP Server & Gateway Updates	124	62.9%	31.0	workflow, add, test
6	Issue & Workflow Management	111	51.4%	11.8	bug, workflow, add
5	Agent & Workflow Infrastructure	111	43.2% ⚠️	11.8	add, update, workflow

Key Findings

1. Most Common Task Type: Dependency Updates & Version Bumps (279 tasks, 23.6% of total)

   • This is the largest category of work, showing continuous maintenance and evolution of dependencies

2. Highest Success Rate: Bug Fixes & Error Resolution (74.6% success rate with 114 tasks)

   • The agent excels at targeted bug fixes and error resolution
   • These tasks typically involve focused changes to specific areas of code

3. Lowest Success Rate: Agent & Workflow Infrastructure (43.2% success rate with 111 tasks)

   • Complex infrastructure changes have higher failure rates
   • Many [WIP] PRs indicate experimentation and iterative development
   • Suggests these tasks may benefit from breaking down into smaller changes

4. Average Code Changes: 18.6 files changed per task, 479.8 lines added, 361.5 lines deleted

   • Moderate scope changes across the board

5. Large-Scale Changes: 170 PRs modified >20 files

   • Complex changes spanning many files have higher coordination overhead

Recommendations

Based on clustering analysis:

1. Continue Focus on Bug Fixes & Testing

The agent shows excellent results on bug fixes and test-related tasks (74.6% success rate). These focused, well-scoped tasks play to the agent's strengths.

2. Break Down Infrastructure Changes

Agent & Workflow Infrastructure tasks have the lowest success rate (43.2%). Consider:

• Breaking complex infrastructure changes into smaller, incremental PRs
• Providing more detailed context and requirements for infrastructure tasks
• Adding validation steps or smoke tests for infrastructure changes

3. Optimize Dependency Update Workflows

With 279 dependency update tasks (23.6% of all work), this is a major category. Success rates vary significantly (43.2% to 73.8%) across different types of dependency updates:

• Simple version bumps perform well
• Complex updates with cascading changes need better scoping

4. Consider Breaking Down Large Changes

170 PRs modified >20 files. Large-scale changes could benefit from:

• More granular task decomposition
• Phased rollouts with validation at each step
• Better coordination of multi-file changes

5. Leverage Success Patterns

Tasks with clear scopes and well-defined outcomes (bug fixes, documentation, simple updates) perform best. Apply these patterns to more complex task types.

---

References:

• Analysis Run: §21852146769
• Total PRs Analyzed: 1,182
• Clustering Method: TF-IDF + K-means (k=10, silhouette score: 0.073)
• Data Period: Last 30 days of copilot agent PRs

AI generated by Copilot Agent Prompt Clustering Analysis

• expires on Feb 17, 2026, 5:08 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-17T05:08:40.060Z.

Closed by Workflow