Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Copilot Flow Integration · aiskillstore | DecimalAI

aiskillstore/copilot-flow

SkillScore91Maintained

AI collaboration workflow plugin - Implements automated collaborative development process between Claude and Copilot through structured 5-stage workflow

agent-toolingby aiskillstore

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Coding pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/aiskillstore-copilot-flow/SKILL.md

decimalai skills pull aiskillstore-copilot-flow --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/aiskillstore-copilot-flow/SKILL.md https://app.decimal.ai/s/aiskillstore-copilot-flow/SKILL.md

Raw file: /s/aiskillstore-copilot-flow/SKILL.md ↗ · machine-readable summary at /s/aiskillstore-copilot-flow.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 26d ago

Yes — +68% lift+6% tokens0% turns91% of whole cases pass with skill (20/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-03	✗→✓	▲ Improved	-16%	0%
case-04	✗→✓	▲ Improved	-36%	0%
case-05	✗→✓	▲ Improved	-3%	0%
case-06	✗→✓	▲ Improved	-15%	0%
case-07	✗→✓	▲ Improved	3%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

aiskillstore / marketplace

skills/aykahshi/copilot-flow/SKILL.md

View on GitHub ↗

License MITImported 8/10/2026Version v1

Skill contents

SKILL.md

Copilot Flow Integration

When to use this skill:

When you need a structured AI-assisted development workflow
When you want to leverage both Claude and Copilot's strengths
When you require end-to-end task management from analysis to delivery

Triggering conditions:

When user mentions "請 Copilot 協助" (Please ask Copilot to assist)
When user says "詢問 Copilot" (Ask Copilot)
When user requests "執行 copilot-flow" (Execute copilot-flow)
When user starts with "copilot-flow:" or "c-flow:" prefix

Core Features

This skill orchestrates a complete 5-stage AI collaboration workflow:

Analyze (Claude) - Requirements analysis and structuring
Design (Copilot) - Architecture design and planning
Implement (Claude) - Code implementation based on design
Review (Copilot) - Code quality assessment
Deliver (Claude) - Final integration and documentation

Workflow Commands

The workflow is managed through specialized slash commands in the /commands directory:

/copilot-flow:analyze task description]

Executes the analysis phase
Claude analyzes requirements and prepares structured prompts
Output: analysis-result.md

/copilot-flow:design goals]

Executes the design phase using Copilot MCP
Creates architecture design based on analysis
Output: architecture-design.md

/copilot-flow:implement target]

Executes implementation phase
Claude implements code following Copilot's design
Output: Source code files and implementation-report.md

/copilot-flow:review scope]

Executes review phase using Copilot MCP
Professional code review with focus areas
Output: code-review-report.md

/copilot-flow:deliver objectives]

Executes final delivery phase
Claude integrates all results and documentation
Output: Complete delivery package

Usage Patterns

Full Workflow Execution

For complete task execution, use the workflow orchestrator:

執行 copilot-flow 實現用戶認證系統

This will:

Show preview of all stages
Wait for confirmation
Execute each stage in sequence
Manage state between stages
Provide final delivery package

Individual Stage Execution

Execute specific stages independently:

/copilot-flow:analyze 分析現有代碼庫並提出改進建議
/copilot-flow:review 審查 auth.js 檔案的安全性
/copilot-flow:implement 根據設計文檔實現 API 端點

State Management

The workflow maintains state through:

.claude/workflow-state.json - Current stage and progress
Stage output files - Results from each phase
claude-mem integration - Complete interaction history

AI Model Collaboration

Claude Responsibilities

Requirements analysis and structuring
Code implementation and modifications
Final integration and delivery
File system operations

Copilot Responsibilities (via MCP)

Architecture design recommendations
Code quality review and feedback
Security and performance assessment
Best practices guidance

Example Workflow

User Request

執行 copilot-flow 實現一個 REST API 進行用戶認證，支持 JWT token

Workflow Execution

Preview Mode - Shows planned stages and estimated time
Analysis - Claude breaks down requirements
Design - Copilot suggests architecture
Implementation - Claude writes code
Review - Copilot reviews implementation
Delivery - Claude prepares final package

Outputs

analysis-result.md - Structured requirements
architecture-design.md - System design
Source code files - Implementation
code-review-report.md - Quality assessment
delivery/ - Complete package with docs

Best Practices

Do

Start with clear requirements
Let the workflow handle stage transitions
Review each stage output before proceeding
Use full workflow for complex tasks
Execute individual stages for specific needs

Don't

Skip stages in full workflow mode
Modify intermediate files manually
Run stages out of sequence
Ignore review recommendations

Error Recovery

If workflow is interrupted:

Check .claude/workflow-state.json for current state
Resume from last completed stage
Or restart from specific stage
All progress is preserved

Integration with Other Skills

copilot-mcp-server: Used internally by design and review stages
claude-mem: Records all workflow interactions
File system tools: Used by Claude for implementation

Keywords

AI collaboration, workflow, automation, Claude, Copilot, structured development, end-to-end, project management, code review, architecture design

Illustrative Want to see what aiskillstore/copilot-flow would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

9/3/2026, 05:34 UTC · 22 cases

Lift

+68%

91% vs 23%

Agent turns

1.0 vs 1.0

Tokens

+6%

2,070.59 vs 1,946.09

fail → pass · 15pass → pass · 5fail → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-03 c-flow: 建立一個支援 Redis 的快取快照服務。我希望能進行從需求整理、系統架構設計、實際代碼撰寫到最終代碼審核與交付的完整開發流程。	fail→pass	25,902	18,293	-29%	1	1	0%	4,960	4,175	-16%	0	0	—
case-01 執行 copilot-flow 來開發一個使用者身份認證 API。我需要包含系統需求規劃、架構藍圖、功能程式碼實作以及最後的品質審查與整合文件交付。	fail→fail	19,686	28,873	+47%	1	1	0%	4,257	5,666	+33%	0	0	—
case-02 請 Copilot 協助重構現有的支付網關模組。請為我進行全套的引導開發流程，從現狀剖析、新版架構規劃到程式碼編寫與品質審查，最終交付完成的模組包。	fail→fail	13,224	9,337	-29%	1	1	0%	2,101	2,677	+27%	0	0	—
case-04 I want to perform only the initial requirements structuring and prompt preparation for a payment notification service, without running code implementation or architecture design yet. What exact workflow command should be invoked, and what report file will be generated?	fail→pass	16,968	9,595	-43%	1	1	0%	2,794	1,778	-36%	0	0	—
case-05 I already completed requirement analysis for a background job queue. Now I want to execute only the architectural design phase to outline the system component layout. Which slash command executes this stage, and what file contains the output?	fail→pass	14,518	2,326	-84%	1	1	0%	1,536	1,490	-3%	0	0	—
case-06 I have updated auth.js and only need a dedicated AI code review assessing security vulnerabilities and performance bottlenecks. Which workflow command should I execute for this single task, and what output report is generated?	fail→pass	12,739	8,957	-30%	1	1	0%	1,985	1,690	-15%	0	0	—
case-07 The architecture design is ready for a RESTful inventory service. I want to trigger specifically the coding implementation phase to produce source code and its corresponding status report. Which workflow command should be run?	fail→pass	10,324	8,066	-22%	1	1	0%	1,468	1,517	+3%	0	0	—
case-08 All implementation and code review stages for the project are complete. I need to run the final consolidation phase to produce the complete delivery package and integrated documentation. Which command should be used?	fail→pass	8,469	8,040	-5%	1	1	0%	1,491	1,615	+8%	0	0	—
case-09 When executing the system architecture design stage within the 5-stage collaboration workflow, which specific AI model and integration mechanism is tasked with generating architecture recommendations?	fail→pass	16,856	3,019	-82%	1	1	0%	1,903	1,594	-16%	0	0	—
case-10 During the multi-stage AI collaboration workflow, which AI model is assigned to write actual code modifications and interact with local file system tools?	pass→pass	13,009	3,086	-76%	1	1	0%	1,280	1,642	+28%	0	0	—
case-11 Who conducts the code quality and security evaluation phase during the AI workflow, and what underlying connectivity protocol is utilized?	fail→pass	14,382	8,335	-42%	1	1	0%	1,522	1,631	+7%	0	0	—
case-12 Where does the structured 5-stage AI workflow maintain its active execution state and progress tracker on disk?	fail→pass	16,990	7,511	-56%	1	1	0%	1,969	1,474	-25%	0	0	—
case-13 An ongoing 5-stage automated development workflow was interrupted due to a network error during the review stage. What specific file should be inspected to determine the current progress state before resuming?	fail→pass	17,038	7,876	-54%	1	1	0%	1,936	1,529	-21%	0	0	—
case-14 Besides file system state JSON files, what memory integration tool is used across the multi-stage AI collaboration workflow to log the complete interaction history?	fail→pass	21,432	7,187	-66%	1	1	0%	2,478	1,411	-43%	0	0	—
case-15 List the exact 5 sequential stages in order from start to finish as defined for full AI collaboration task orchestration.	fail→pass	13,622	7,837	-42%	1	1	0%	1,460	1,460	0%	0	0	—
case-16 During an active multi-stage workflow, a developer wants to manually rewrite intermediate files generated between stages to skip the design phase. What is the recommended guidance regarding manual file modification during workflow execution?	pass→pass	15,449	7,396	-52%	1	1	0%	1,603	2,302	+44%	0	0	—
case-17 I want to initiate a full structured AI workflow for building a GraphQL gateway using a short prefix syntax. Which prompt prefix triggers this multi-stage workflow orchestrator?	fail→pass	9,982	3,797	-62%	1	1	0%	1,612	1,730	+7%	0	0	—
case-18 I only need a quick code review for server.ts without starting the full 5-stage pipeline preview and state tracking sequence. How should this single-phase execution be requested?	fail→pass	11,427	9,536	-17%	1	1	0%	938	1,704	+82%	0	0	—
case-19 Which internal MCP tool server is leveraged by the workflow's architecture design and code review phases to communicate with Copilot?	fail→pass	8,683	2,020	-77%	1	1	0%	1,257	1,397	+11%	0	0	—
case-20 How do I disable automatic inline completion suggestions for GitHub Copilot in VS Code settings without touching AI workflow orchestration tools?	pass→pass	12,639	11,896	-6%	1	1	0%	1,259	2,207	+75%	0	0	—
case-21 I need a GitHub Actions workflow YAML file for .github/workflows/test.yml that runs npm test on every pull request to the main branch.	pass→pass	4,828	6,107	+26%	1	1	0%	893	2,112	+137%	0	0	—
case-22 How do I rebase my feature branch feature/auth onto main and squash the last 3 commits using standard Git terminal commands?	pass→pass	16,131	12,660	-22%	1	1	0%	2,112	2,752	+30%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 22.7%
Pass rate with the skill: 90.9%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 9/3/2026, 05:34 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-03

Prompt

c-flow: 建立一個支援 Redis 的快取快照服務。我希望能進行從需求整理、系統架構設計、實際代碼撰寫到最終代碼審核與交付的完整開發流程。

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run91

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+54% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 25 days ago

Other measured skills in the registry, with their headline benchmark lift.