Subtasks and blockers were dropped in migrations 008 and 011, and
queued became ready in 012; the stale feature list is why agents kept
reaching for a `ranger blocker` command.

Assisted-by: Claude Opus 5 via Claude Code

diff --git a/README.md b/README.md
index 4178380..1c20cc5 100644
--- a/README.md
+++ b/README.md
@@ -16,10 +16,10 @@ Every item is a task. No stories, bugs, chores, or features — just tasks. Each
 
 - **Title** and **description**
 - **Comments** for ongoing notes
-- **Subtasks** for breaking work down
-- **Blockers** for expressing dependencies
 - **Tags** for grouping related work
 
+Ranger has no subtasks and no dependency edges. Both were built and then removed as complexity that went unused, leaving a flat list ordered by priority. Break work down by creating separate tasks; record a dependency as a comment.
+
 ### States
 
 A task is always in one of four states:
@@ -27,7 +27,7 @@ A task is always in one of four states:
 | State | Meaning |
 |---|---|
 | **Icebox** | Captured but not committed to |
-| **Queued** | Committed and ordered by priority |
+| **Ready** | Committed and ordered by priority |
 | **In Progress** | Actively being worked on |
 | **Done** | Finished |
 
@@ -37,7 +37,7 @@ Tags replace projects. Filter any backlog by tag to see a focused slice of work.
 
 ### Interface
 
-The webapp uses an expanding modal for editing tasks — no cluttered inline editing. Design is minimalist, built with [Utopia](https://utopia.fyi/) fluid responsive CSS.
+`ranger serve` starts a web server on port 3000 that renders a backlog as a board. Tasks in the ready and icebox columns drag to reorder; everything else is read-only for now. Design is minimalist, built with [Utopia](https://utopia.fyi/) fluid responsive CSS.
 
 ## Getting Started
 
@@ -58,8 +58,9 @@ cargo run --bin ranger -- <command>
 ```
 ranger backlog create "my-project"
 ranger task create "First thing to do" --backlog my-project
-ranger task create "Second thing" --backlog my-project --state ready --tag urgent
+ranger task create "Second thing" --backlog my-project --state ready
 ranger task list --backlog my-project
+ranger tag add <key> urgent
 ranger task edit <key> --state in_progress
 ranger comment add <key> "Started working on this"
 ranger task show <key>
@@ -71,13 +72,17 @@ Set `RANGER_DEFAULT_BACKLOG` to skip `--backlog` on every command.
 
 The database lives at `$XDG_DATA_HOME/ranger/ranger.db` by default. Override with `--db <path>` or `RANGER_DB` env var.
 
+### For Claude Code
+
+This repository doubles as a Claude Code plugin marketplace. Add it with `/plugin marketplace add kejadlen/ranger`, then install the `ranger` plugin to give an agent [the skill](skills/ranger/SKILL.md) that documents this workflow.
+
 ## Architecture
 
 Ranger is a single Rust crate with a library and binary target:
 
 - **Library** (`ranger`) — core data model, database operations, key generation
 - **CLI** (`ranger` binary) — clap-based binary for humans and AI agents
-- **Webapp** — for human use (planned)
+- **Webapp** (`ranger serve`) — axum server in the same binary, serving the board
 
 The CLI exists primarily so AI agents can manage tasks programmatically. The webapp exists for humans who prefer a visual interface.
 
@@ -92,11 +97,11 @@ The version format is calver: the date of the commit plus its short SHA (e.g., `
 
 ## Roadmap
 
-**First milestone:** self-host Ranger so an AI agent can use it for task management while building Ranger itself.
+The first milestone is met: Ranger's own backlog lives in Ranger, and an agent works it while building the tool. The Claude Code plugin has shipped too.
 
-After that:
+Next:
 
-- Claude Code plugin for AI agent task management
+- Configurable board columns in the web UI
+- Auto-refreshing the board when data changes
 - Public read-only sharing with permissions
-- Configurable backlog views
 - Browsing completed work