Cross Platform Handoff

Cross-Platform Handoff Patterns

Guide for seamless handoff between Claude Code and GitHub Copilot agent environments when performing document accessibility audits.

Platform Comparison

Capability	Claude Code	GitHub Copilot	Gemini CLI	Codex CLI
Agent format	.claude/agents/*.md	.github/agents/*.agent.md	.gemini/extensions/*/skills/*/SKILL.md	.codex/AGENTS.md
Frontmatter	maxTurns, memory	tools, agents, model, handoffs	name, description	N/A (inline)
Sub-agents	Parallel via Task tool	Via agent tool + agents frontmatter	Via skill references	N/A
Hidden helpers	N/A (all agents user-invokable)	user-invokable: false in frontmatter	N/A	N/A
Skills	N/A	.github/skills/*/SKILL.md	Extension skills	N/A
Context file	CLAUDE.md	copilot-instructions.md	GEMINI.md	AGENTS.md
Interactive UI	Terminal prompts	askQuestions tool	Terminal prompts	Terminal prompts

Handoff Scenarios

Scenario 1: Start in Claude Code, Continue in Copilot

When: You start a document audit in the terminal with Claude Code but need to continue editing or remediation in VS Code.

Steps:

1. In Claude Code, run the document-accessibility-wizard
2. The wizard generates DOCUMENT-ACCESSIBILITY-AUDIT.md
3. Open the project in VS Code
4. In Copilot Chat, select @document-accessibility-wizard
5. Say: "Continue the audit from the existing report"
6. The Copilot agent reads the existing report and picks up where Claude left off

What transfers automatically:

• The audit report file (DOCUMENT-ACCESSIBILITY-AUDIT.md)
• Scan configuration files (.a11y-office-config.json, .a11y-pdf-config.json)
• SARIF output (if generated)

What does NOT transfer:

• In-memory session state (conversation history)
• Partial scan progress (if interrupted mid-scan)

Scenario 2: Start in Copilot, Continue in Claude Code

When: You start an audit in VS Code but need CLI-level access for batch operations.

Steps:

1. In Copilot Chat, run @document-accessibility-wizard
2. Generate the audit report
3. Switch to terminal with Claude Code
4. Say: /document-accessibility-wizard Re-scan with comparison against the existing report
5. Claude Code reads the previous report and runs a delta comparison

Scenario 3: Parallel Platform Scanning

When: Large document libraries benefit from splitting work across platforms.

Steps:

1. In Claude Code: scan PDFs (better for binary parsing in terminal)
2. In Copilot: scan Office documents (better with VS Code file explorer integration)
3. Both write to separate report files
4. In either platform: "Compare and merge these two audit reports"

Shared Artifacts

These files are platform-agnostic and work identically in both environments:

Artifact	Path	Purpose
Audit report	DOCUMENT-ACCESSIBILITY-AUDIT.md	Main audit output
Office config	.a11y-office-config.json	Scan rule configuration
PDF config	.a11y-pdf-config.json	PDF scan rules
SARIF output	*.sarif.json	CI/CD integration
Scan templates	templates/*.json	Profile presets
Batch scripts	*.ps1 / *.sh	Remediation automation

Report Format Compatibility

Both platforms generate reports in the same markdown format with identical sections:

1. Audit Information
2. Executive Summary
3. Accessibility Scorecard
4. Cross-Document Patterns
5. Findings by File
6. Remediation Priority
7. Confidence Summary
8. Template Analysis
9. Comparison Report (if re-scan)

This means any report generated by one platform can be:

• Read and continued by the other platform
• Used as a baseline for comparison scans on either platform
• Parsed by CI/CD pipelines regardless of which platform generated it

Agent Skills as Shared Knowledge

The Agent Skills in .github/skills/ serve as platform-independent knowledge:

• document-scanning - File discovery patterns, delta detection commands
• accessibility-rules - Rule reference with WCAG 2.2 mapping
• report-generation - Scoring formulas, report structure, compliance formats

Claude Code agents access this knowledge through their agent instructions. Copilot agents access it through the Skills framework (auto-resolved from SKILL.md descriptions).