Cloud Migration AI Repo

GitLab Duo skills, prompts, rules and tools for a multi-application cloud migration. Everything runs in GitLab Duo Chat in the IDE from files in this repo. No AI Catalog is needed.

Catalog

Asset	Type	Migration phase	Status	What it does
/functional-doc	Skill: one command	Discover & assess	Draft v0.1	/functional-doc docs/confluence → code scan → prepare docs (incl. OCR) → functional doc → Word + Excel
/app-functionality-extractor	Skill (step 1)	Discover & assess	Draft v0.1	Scans the app repo → docs/functional-spec.md (design notes: gitlab-duo-app-functionality-agent.md)
/app-functional-doc-builder	Skill (step 2)	Discover & assess	Draft v0.1	Code spec + Confluence (incl. diagram notes) → Functional Document → docs/functional-doc.md/.json + Excel + Word (docs)
/migration-requirements-extractor	Skill (step 3)	Discover & assess → Plan waves	Draft v0.1	Reviewed functional doc → traceable on-prem → cloud requirements → Excel (docs)
confluence-app-extractor	Agent prompt (copy-paste)	Discover & assess	Draft v0.1	Fills the application template page from existing Confluence pages, with citations
doc-prep	Tool	Discover & assess	Tested on samples	Confluence exports (Word/PDF/HTML/Excel/images/draw.io) → text; OCR on images; redacts secrets
install-to-app-repo.ps1	Tool	All	Tested	Copies the skills + tools into a cloned app repo
office-export	Tool / CI job	All	Tested on samples	Validates the skills' JSON and exports Excel + Word
how-to-build-app-functional-doc	Confluence playbook	Discover & assess	Draft	Step-by-step guide for the three skills

How to use it (per app)

Full end-to-end workflow with diagram: WORKFLOW.md.

Requirements (verify on your GitLab version): GitLab 18.10+, Premium/Ultimate with Duo Agent Platform, GitLab for VS Code 6.71.4+, Python 3.10+, Windows 10/11 (for the built-in OCR). Docs: https://docs.gitlab.com/user/duo_agent_platform/customize/agent_skills/

#	Step	How
1	Clone the app repo	git clone <app-repo-url> C:\work\<app>
2	Add the AI toolkit to it	From this repo: .\tools\install-to-app-repo.ps1 -AppRepo C:\work\<app> -InstallPython (copies .agents/skills, tools/doc-prep, tools/office-export)
3	Add the Confluence documents	Export each page (⋯ → Export → Word, or HTML with attachments) into docs/confluence/. Add draw.io diagrams as .drawio, plus images. Optional: <image>.image.md describing key diagrams.
4	Run one command	VS Code → GitLab Duo Chat → new conversation → /functional-doc docs/confluence
5	Review	SME reviews docs/out/<app>-functional-doc.docx/.xlsx; give the answers back in the chat to update
6	Requirements	New conversation → /migration-requirements-extractor Cloud: <..>. Pattern: <..>. → docs/out/<app>-migration-requirements.xlsx
7	Commit	Commit docs/ on a branch and open an MR

What /functional-doc does: Stage 1 scans the code → docs/functional-spec.md (reused next time, add RESCAN to redo) · Stage 2 runs tools/doc-prep/prepare_sources.py → text + OCR of images → docs/confluence/_extracted/ · Stage 3 builds the functional doc from both → docs/functional-doc.md/.json · Stage 4 exports Word + Excel to docs/out/.

The individual skills (/app-functionality-extractor, /app-functional-doc-builder) still work on their own when you need to re-run one stage. After you edit a SKILL.md, re-run the install script and start a new Duo conversation.

Confluence via MCP (optional, instead of exports): see ide/. Exports are recommended because they include images and work offline.

Layout

.
├── .agents/skills/<skill>/SKILL.md   # what Duo runs (source of truth for instructions)
├── agents/<name>/                    # design notes, evals, extra prompts for each skill/agent
├── tools/doc-prep/                   # Confluence exports → text + OCR of images
├── tools/office-export/              # JSON → Excel/Word + validation (+ CI job snippet)
├── tools/install-to-app-repo.ps1     # copies skills + tools into an app repo
├── confluence/                       # Confluence page drafts (playbooks, guides)
├── ide/                              # IDE config (mcp.json for Confluence Cloud)
└── gitlab-duo-app-functionality-agent.md

Adding a new skill

1. Create .agents/skills/<kebab-case-name>/SKILL.md with frontmatter name (same as the folder), description (what it does and when to use it, because Duo picks skills by this) and metadata: slash-command: enabled.
2. Put design notes and evals in agents/<name>/.
3. Add a row to the Catalog.
4. Run the evals on 2–3 real apps before telling the team.