feat: bundle review agents (critic + planner + reviewer) + bump to 0.9.0

[Evgenii-Konev]

What

Bundles three read-only, max-effort Opus subagents in the plugin (agents/, auto-discovered, invokable as projectstore:<name>), and wires the critic into the peer-review flow.

Agent	When	Role
projectstore-critic	After authoring/revising an ADR / research / epic / story or a design proposal	Adversarial, fresh-context critic — pre-commit to likely problems, verify claims against source, rate assumptions, gap-analysis + pre-mortem. Verdict ship / revise / rethink.
code-planner	Before writing non-trivial code	Returns where/how a change fits this repo (placement, reuse, conventions, pitfalls, ordered plan) — grounded, not generic.
code-reviewer	After writing code, before committing	Spec compliance → correctness → regressions → codebase-fit → tests, severity + confidence rated, self-audited. Verdict commit / fix first.

Why

projectstore already spawns an ad-hoc critic in /projectstore:review, but relied on oh-my-claudecode:critic / general-purpose. This ships a purpose-built critic with the plugin so the review channel works out of the box — "used by projectstore always, especially the critic."

Changes

• agents/ — three new agent files (projectstore-critic, code-planner, code-reviewer). model: opus, effort: max, read-only tool sets.
• commands/review.md — /projectstore:review now prefers projectstore:projectstore-critic, falling back to oh-my-claudecode:critic then general-purpose.
• README.md — new Bundled agents section (English) + inventory/What's in the box entry.
• plugin.json / marketplace.json — version 0.8.0 → 0.9.0.

Notes

• Agent descriptions/bodies are in English to match the repo.
• code-reviewer was generalized — internal inference-codebase names (vLLM / project-specific test targets) were removed so the agent is repo-agnostic for public distribution.
• Plugin agents resolve at the lowest priority, so a same-named agent in ~/.claude/agents/ or .claude/agents/ transparently overrides the bundled one — no collision for users who already run their own.

https://claude.ai/code/session_01Rm7k8HUn3KohZgcD8XTyCC