llm-wiki-migration-planner

name	llm-wiki-migration-planner
description	Create a no-write migration plan for turning existing notes, docs, Obsidian vaults, Notion exports, research archives, or repository documentation into an LLM-Wiki. Use when the user wants a dry-run plan before moving or editing files; route applying the approved plan to llm-wiki-refactor.
license	MIT
compatibility	Designed for Agent Skills-compatible coding agents. Requires read access only; applying the plan should be delegated to llm-wiki-refactor.
metadata	{"author":"po4yka","version":"1.0.0","install_scope":"self-contained"}

LLM-Wiki Migration Planner

Goal

Produce a safe, reviewable migration plan without changing the user's files.

When to use

• The user wants a plan before migrating notes, docs, or an Obsidian vault into an LLM-Wiki, without touching any files yet.
• The user is deciding between a personal vault, repo wiki, team wiki, research corpus, or product prototype target and needs a mapping proposal first.
• The source is mixed content (Notion export, chat export, research PDFs, company docs) and needs profile-specific classification before any move happens.
• The user asks for a dry-run, review, or approval step ahead of running llm-wiki-refactor.

Inputs

• Current document root.
• Desired target: personal vault, repo wiki, team wiki, research corpus, or product prototype.
• Existing constraints: Obsidian, Notion export, docs generator, git, privacy, attachments.

Procedure

1. Confirm dry-run mode

State that this skill produces a plan only. Do not move, rename, delete, normalize frontmatter, or rewrite links.

2. Identify migration profile

Classify the source:

Profile	Notes
Obsidian vault	Preserve wikilinks, aliases and attachments.
Repository docs	Preserve README/ADR/docs generator conventions.
Notion export	Normalize nested pages and attachment paths.
Readwise/Omnivore export	Treat exports as raw sources first.
Chat exports	Separate raw conversation from reusable query pages.
Research PDFs	Preserve PDFs in raw/assets or raw/sources.
Company docs	Require PR-based write model and permissions.

3. Inventory and classify

For each candidate file, infer:

• current path;
• likely page type;
• whether it is raw, wiki, synthesis, query, asset, or mixed;
• confidence;
• migration risk;
• proposed target path.

4. Build mapping table

Use this table:

| Current path | Proposed path | Type | Confidence | Risk | Notes |
|---|---|---|---:|---|---|

Use confidence values conservatively. Low-confidence rows should require human review. Re-verify each row's proposed path and confidence against the actual current file contents, not just filenames, before finalizing the table.

5. Design phases

Split the migration into batches:

1. create scaffold;
2. preserve raw sources and assets;
3. move source summaries;
4. create or update index/log;
5. normalize frontmatter;
6. repair links;
7. run lint and trust audit.

6. Produce apply instructions

Give exact next steps for llm-wiki-refactor, but do not execute them.

Output

## Migration thesis

## Source profile

## Target structure

## Migration table

## High-risk files

## Phased plan

## Required approvals

## Next skill

Safety gates

• No file writes.
• No deletion recommendations without a separate archive/backup option.
• Do not convert Obsidian wikilinks to Markdown links unless explicitly requested.
• Do not collapse raw sources into summaries.
• Do not propose cloud ingestion for sensitive material without explicit approval.