instructa · regenrek · Jul 20, 2026 · Jul 19, 2026
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -185,6 +185,7 @@ jobs:
         with:
           node-version: 24
           registry-url: https://registry.npmjs.org
+          package-manager-cache: false
 
       - name: Download release assets
         env:

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -11,15 +11,16 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 ### Fixed
 
 - Locked the client provisioning contract in CLI help, generated reference, docs, and E2E tests: Codex installs project MCP/hooks but no project roles or skills; Claude Code installs project MCP/standalone roles/hooks while its plugin owns skills; Cursor installs project MCP/roles/skills/hooks. `--no-mcp` skips only MCP, and `--no-hooks` independently skips hooks.
+- Removed the current routing-bundle application surface from Planr docs, generated references, tests, and release wording. Planr now documents only provider-neutral repository declarations and route evidence; optional routing lifecycle is external and repository-local, with Switchloom v0.2.0 as an external tool rather than something Planr invokes, installs, or uninstalls.
 
 ## [1.5.0] - 2026-07-17
 
-Routing policy becomes an optional package instead of a responsibility compiled into Planr Core. Planr keeps the provider-neutral declaration, evidence, and safe repository transaction boundary; `planr-routing` owns volatile model and host knowledge.
+Routing policy became an optional package instead of a responsibility compiled into Planr Core. This section is historical: the routing-bundle application boundary described here was removed in 1.5.1.
 
 ### Changed
 
-- Hard-cut opinionated model routing out of Planr Core. Core now owns only the provider-neutral registry, route resolution and evidence, plus strict RoutingBundle v1 inspect/preview/apply. The independently buildable `planr-routing` workspace package owns named policies, exact models and effort, Codex/Claude Code/Cursor/mixed-host bindings, generated roles and skills, evaluation, signing, registry data, and catalog publication.
-- Removed the legacy preset CLI, MCP tools, Rust modules, root policy fixtures, and root website ownership without aliases or compatibility layers. The replacement flow compiles a bundle with `planr-routing` and applies it through `planr routing bundle`; no command edits user configuration.
+- Hard-cut opinionated model routing out of Planr Core. At the time, Core owned the provider-neutral registry, route resolution and evidence, plus strict RoutingBundle v1 inspect/preview/apply. The independently buildable `planr-routing` workspace package owned named policies, exact models and effort, Codex/Claude Code/Cursor/mixed-host bindings, generated roles and skills, evaluation, signing, registry data, and catalog publication.
+- Removed the legacy preset CLI, MCP tools, Rust modules, root policy fixtures, and root website ownership without aliases or compatibility layers. The historical replacement flow compiled a bundle with `planr-routing` and applied it through `planr routing bundle`; no command edited user configuration.
 - Catalog entries remain deterministic, experimental, and unrecommended. Offline or caller-asserted evidence cannot promote them. Detached signatures and signed bundles require an independently supplied trusted signer and Ed25519 public key.
 - Hardened repository application against parent/child artifact collisions and rollback residue. Global Codex, Claude Code, Cursor, shell, keychain, credential, and XDG sentinels remain unchanged across bundle application.
 
@@ -34,7 +35,7 @@ Verified presets turn model routing from a hand-authored host configuration into
 - Reproducible preset evaluation through `planr agents preset evaluate`: versioned challenge tasks, Planr-read artifact hashes, task-bound outcome oracles, lifecycle thresholds, optional live-host execution, and independently pinned Ed25519 telemetry receipts. Recommendations require complete, current, trusted route and usage evidence; offline estimates and incomplete runs cannot recommend.
 - Optional signed preset registry commands (`verify`, preview-first `import`, and offline `list`) with immutable manifest-hash-addressed caching, lifecycle and compatibility checks, separately provisioned maintainer trust, and re-verification of cached content. Active projects and previously imported packs continue working when the registry is unavailable.
 - Public [Planr Preset Catalog](https://planr-test-catalog.office-35d.workers.dev/) generated from the canonical verifier and evaluation report. Repository-owned Alchemy/Cloudflare tooling builds an allowlisted static publication, deploys an isolated `test` stage, and ships restrictive response headers without storing private signing keys in the site or deployment environment.
-- Detailed guides for [preset composition](planr-routing/docs/PRESET_COMPOSITION.md), [evaluation](planr-routing/docs/PRESET_EVALUATION.md), and the [registry](planr-routing/docs/PRESET_REGISTRY.md), plus CLI, MCP-contract, architecture, and deployment documentation.
+- Historical guides covered preset composition, evaluation, and the registry, plus CLI, MCP-contract, architecture, and deployment documentation.
 
 ### Security
 

diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -19,7 +19,6 @@ path = "src/main.rs"
 anyhow = "1"
 clap = { version = "4", features = ["derive"] }
 cliclack = "0.5.4"
-ed25519-dalek = "2"
 rusqlite = { version = "0.32", features = ["bundled"] }
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
@@ -49,7 +48,3 @@ tempfile = "3"
 lto = "thin"
 codegen-units = 1
 strip = "symbols"
-
-[workspace]
-members = ["planr-routing"]
-resolver = "2"
diff --git a/README.md b/README.md
@@ -54,7 +54,7 @@ Manual downloads, from-source builds, and client wiring details: [Install Guide]
 
 ## Install The Plugin (Skills)
 
-The plugin under `plugins/planr` carries the ten Planr workflow skills. Optional model-routing roles come from repository-local routing bundles. The `planr` CLI (above) is required separately.
+The plugin under `plugins/planr` carries the ten Planr workflow skills. Optional model-routing declarations live in repository-local files such as `.planr/agents.toml` and `.planr/policy.toml`; external tools may manage those files, but Planr does not install or invoke a routing engine. The `planr` CLI (above) is required separately.
 
 <a id="install-plugin-codex"></a>
 <details>
@@ -112,7 +112,7 @@ planr prompt cli
 
 ## Tell Your Agent
 
-Remember one public entry point: `$planr`. It routes the request from live Planr state, including planning-only work and autonomous-goal preparation. Stage skills are advanced surfaces selected by the router.
+Remember one public entry point: `$planr`. It routes ordinary planning and status work from live Planr state. For long autonomous runs, use the explicit two-step workflow: `$planr-goal` prepares durable state, then `$planr-loop` executes the resulting plan.
 
 Start a new product from an idea:
 
@@ -126,7 +126,7 @@ split an MVP build plan, check it, then build the Planr map. Do not implement ye
 For a long autonomous run, prepare outside the driver first. The preparation result prints a real plan id; Codex or Claude Code then starts only the plan-bound loop driver:
 
 ```text
-Use $planr to prepare an autonomous goal for the weekly overview feature.
+Use $planr-goal to prepare an autonomous goal for the weekly overview feature.
 
 /goal Use $planr-loop on plan <plan-id>. The loop contract is stored in planr
 context (tag: goal-contract).
@@ -140,7 +140,7 @@ Mid-project work (a new feature, refactor, or fix on an existing project) works
 
 ## What's new
 
-- **1.5.0 — Optional routing policies:** Planr Core stays provider-neutral while the independently buildable `planr-routing` package owns model policies and host bindings for Codex, Claude Code, Cursor, and mixed-host setups. Start with [Routing Bundles](docs/ROUTING_BUNDLES.md), the [`planr-routing` guide](planr-routing/docs/MODEL_ROUTING_POLICY.md), and the [1.5.0 release notes](https://github.com/instructa/planr/releases/tag/v1.5.0).
+- **1.5.1 — External routing boundary:** Planr consumes provider-neutral repository declarations and route evidence only. Optional model-routing lifecycle is external, for example [Switchloom v0.2.1](https://github.com/instructa/switchloom/releases/tag/v0.2.1) managing repository-local files outside Planr. Start with [Model Routing](docs/MODEL_ROUTING.md), [Switchloom](https://switchloom.ai), the [Switchloom repository](https://github.com/instructa/switchloom), its tagged [setup quickstart](https://github.com/instructa/switchloom/blob/v0.2.1/README.md#setup-from-the-website) and [lifecycle docs](https://github.com/instructa/switchloom/blob/v0.2.1/docs/preset-composition.md#repository-lifecycle-commands), and the [Changelog](CHANGELOG.md).
 - **1.4.0 — Verified presets:** Added policy-driven composition, evaluation, signed registry evidence, and the public catalog. See the [1.4.0 release notes](https://github.com/instructa/planr/releases/tag/v1.4.0).
 - **1.3.0 — Native host hooks:** Added automatic session-state injection and loop recovery for supported hosts. See the [Hooks guide](docs/HOOKS.md) and [1.3.0 release notes](https://github.com/instructa/planr/releases/tag/v1.3.0).
 

diff --git a/apps/docs/content/docs/agents/prompt-recipes.mdx b/apps/docs/content/docs/agents/prompt-recipes.mdx
@@ -3,7 +3,7 @@ title: Agent Prompt Recipes
 description: Copy accurate prompts for planning, autonomous preparation, status, recovery, and advanced direct routing.
 ---
 
-Use `$planr` for normal work. These prompts describe intent and boundaries; the router reads live Planr state and chooses the correct stage. Replace “my request” with the outcome and proof that matter to you.
+Use `$planr` for normal work. Autonomous delivery is the exception: prepare with `$planr-goal`, then execute only the generated `$planr-loop` handoff. Replace “my request” with the outcome and proof that matter to you.
 
 ## First request
 
@@ -51,7 +51,7 @@ Recovery begins with map, lease, review, approval, log, and preview evidence. It
 
 ## Advanced direct entry points
 
-Direct stage skills are optional precision controls for experienced users. They do not replace `$planr` as the documented default.
+Direct stage skills are precision controls for experienced users. They do not replace `$planr` as the documented default for ordinary work; autonomous runs intentionally use `$planr-goal` followed by `$planr-loop`.
 
 ### `$planr-plan`: plan a bounded scope directly