feat(skills): add v2 data refactoring skills (migrator, data-api, renderer)

[DeJeune]

What this PR does

Before this PR:

No structured guidance existed for Claude agents working on the v2 data layer migration. Each migration task required re-explaining the architecture, patterns, and constraints from scratch.

After this PR:

Three new skills guide the full v2 data refactoring lifecycle:

• v2-migrator (Phase 1): Moving legacy Redux/Dexie/ElectronStore data into SQLite — covers source classification, migrator contract, cross-domain references & FK design, orphan handling, and streaming patterns
• v2-data-api (Phase 2): Building Main-process services — covers Handler→Service→Repository layered architecture, API schema design, preference schema, and database patterns
• v2-renderer (Phase 3): Wiring up Renderer-process hooks — covers useQuery/useMutation (DataApi), usePreference (settings), and useCache/useSharedCache/usePersistCache (temporary data) with cache tier selection guidance and multi-window sync considerations

All three skills enforce strict TDD (red-green-refactor).

Why we need it and why it was done in this way

The v2 data refactoring involves migrating ~391 classified data items across Redux, Dexie, and ElectronStore into a unified SQLite + Preference + Cache architecture. Without structured skills, agents repeatedly make mistakes around:

• Cross-domain references: Redux stores full object copies (e.g., Model, KnowledgeBase[]) instead of IDs, leading to stale phantom data when originals are deleted. Skills document the FK-IDs-only pattern and orphan reference handling.
• Cross-category migration: Some preferences live in domain slices (memory.embedderModel, websearch.compressionConfig.embeddingModel) rather than settings. Skills guide correct reclassification.
• Cache tier selection: Three cache tiers (memory/shared/persist) have different scope, sync, and persistence characteristics. Skills provide a decision flow.

The following tradeoffs were made:

Skills are split into 3 phases (migrator → data-api → renderer) matching the natural dependency order, rather than one monolithic skill.

The following alternatives were considered:

A single combined skill was considered but rejected — each phase has distinct concerns, file locations, and testing strategies.

Breaking changes

None. This PR only adds new skill files and updates .gitignore/public-skills.txt.

Special notes for your reviewer

• All skills passed pnpm skills:check (8 public skills validated)
• The skills reference existing documentation in docs/en/references/data/ and existing migration code in src/main/data/migration/v2/
• The cross-domain reference tables in v2-migrator were built from actual Redux store analysis (src/renderer/src/store/)

Checklist

• PR: The PR description is expressive enough and will help future contributors
• Code: Write code that humans can understand and Keep it simple
• Refactor: You have left the code cleaner than you found it (Boy Scout Rule)
• Upgrade: Impact of this change on upgrade flows was considered and addressed if required
• Documentation: A user-guide update was considered and is present (link) or not required. Check this only when the PR introduces or changes a user-facing feature or behavior.
• Self-review: I have reviewed my own code (e.g., via /gh-pr-review, gh pr diff, or GitHub UI) before requesting review from others

Release note

NONE

[0xfullex]

Note

This issue/comment/review was translated by Claude.

Pay attention to regular maintenance

---

Original Content

注意定期维护