GQAdonis · Sep 26, 2026 · Sep 26, 2026
diff --git a/README.md b/README.md
@@ -33,6 +33,7 @@ flow, sequence, and C4 diagrams throughout. This README is the quick tour; the g
 | Understand the loops | [Loop Architecture](docs/guide/03-loop-architecture.md) · [Four-Layer Pipeline](docs/guide/04-four-layer-pipeline.md) |
 | Know the substrate | [MCP Servers](docs/guide/05-mcp-substrate.md) · [Memory & Learning](docs/guide/06-memory-and-learning.md) · [Sycophancy Correction](docs/guide/07-sycophancy-correction.md) |
 | Browse every skill | [Skills Overview](docs/guide/08-skills-overview.md) · [Process Skills](docs/guide/09-process-skills.md) · [Language Skills](docs/guide/10-language-skills.md) · [Artifact Refiner](docs/guide/11-artifact-refiner.md) · [Native Agent Generator](docs/guide/12-native-agent-generator.md) |
+| Route UI work | [UI/UX routing guide](docs/guide/25-ui-ux-routing.md) · [Command reference](docs/reference/ui-ux-routing.md) |
 | Coordinate agents | [Agent teams guide](docs/guide/24-agent-teams.md) · [Requests, exports and ownership](docs/agent-teams.md) |
 | Reference the engine room | [Tools](docs/guide/13-tools-reference.md) · [Rust Toolchain](docs/guide/14-rust-toolchain.md) · [Hooks & Lifecycle](docs/guide/15-hooks-and-lifecycle.md) · [CLI & Scripts](docs/guide/16-cli-and-scripts.md) |
 | Install, run, contribute | [Platform Support](docs/guide/17-platform-support.md) · [Plugins & Marketplace](docs/guide/18-plugins-and-marketplace.md) · [Installation](docs/guide/19-installation.md) · [Updating](docs/guide/20-updating.md) · [Contributing](docs/guide/21-contributing.md) |
@@ -107,6 +108,25 @@ service. See the [guide](docs/guide/24-agent-teams.md) for the novice flow and
 [request reference](docs/agent-teams.md) for exact JSON examples, model and
 memory limits, and the distinction between skill distribution and native teams.
 
+## UI/UX routing
+
+Use `prometheus-ui-ux` for UI, interaction, tokens, motion and on-screen copy. It
+selects project context, focused Pro Max data, craft and platform guidance while
+preserving design authority. New work and authorized redesign may use taste;
+refinement and review do not. The catalog has 41 entries, including 40 portable
+entries shared with mini. Full-only Impeccable engine execution requires a
+preinstalled `IMPECCABLE_BIN`; the portable core does not claim engine parity.
+
+Adopt existing teams through creator `install-project`: it preserves an explicit
+selection, adopts a sole team and requires a choice when ambiguous. Relevant team
+roles then apply to all code work; UI guidance activates only for UI work. Existing
+native configuration stays intact, and unavailable delegation uses a disclosed
+sequential fallback. `export` remains proposal-only.
+
+The completed-phase helper returns an evidence checklist, not a review result.
+See the [guide](docs/guide/25-ui-ux-routing.md) for commands, independent review and
+the remaining Windows, live-harness, Electron and release-certification gaps.
+
 ## Native Agent Generator
 
 The skill pack includes `/create-native-agent` — a project scaffold that generates

diff --git a/docs/agent-teams.md b/docs/agent-teams.md
@@ -109,6 +109,25 @@ Claude and Kimi have verified agent plugin/marketplace exports. Install either t
 
 UAR defaults require explicit review of tool policy and bundles before registration; skill preference is not a deny policy. BossFang's native `skills=[]` means all unless `skills_disabled=true`; an empty portable role skill list exports the disabled form. Supply an approved service URL and credential reference for registration, retain returned IDs, and account for partial success. Hand activation can start autonomous schedules. Neither export nor discovery establishes authorization to mutate a service.
 
+## Install or adopt a project team
+
+After reviewing the export, finish normal project-team creation with `install-project`. For a new team, save `{"project":"/path/to/project","team":<manifest>}` as `install-request.json`, then run:
+
+```text
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --input install-request.json --dry-run
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --input install-request.json
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --project "/path/to/project" --check
+```
+
+For an existing team, use `--project "/path/to/project"` without an input manifest. A recorded `.agent-team/project-routing.json` selection wins; otherwise a sole `.agent-team/<id>/team.json` is adopted. Multiple candidates require `--team <id>`; stale selections fail rather than silently switching. Intentional manifest replacement requires `updateTeam: true` in the request. Creator check exits 2 for drift and 1 for errors.
+
+Installation writes managed discovery pointers to both instruction entrypoints, the active routing record and missing native definitions. Existing native files, role IDs, ownership, model policies, permissions and concurrency remain intact; differing configuration is reported for deliberate merge. Recovery records retain prior instruction bytes. Export stages proposals; installation establishes discovery. Neither starts execution or activates UAR/BossFang registration.
+
+All code tasks use the selected team’s relevant roles. UI roles conditionally use `prometheus-ui-ux`; UI review uses `prometheus-ui-review` after the whole implementation phase, without taste or user-only skill preloads. Backend work loads no UI guidance. Use native delegation only if available, otherwise disclose sequential role execution; builder self-review is not independent review. Zed receives the pointer in its effective existing instruction file. External ACP agents keep native configuration, and Zed parallel threads are not a delegation API.
+
+See [UI/UX routing](guide/25-ui-ux-routing.md) for the selective workflow.
+
+
 ## Choose models from evidence
 
 `models-discover` accepts `kind: "openai"` with a configured `baseUrl`, or `kind: "uar"`/`"bossfang"` with an explicit `discoveryUrl`. Authentication uses `auth.env` referencing a local environment variable; do not embed tokens in JSON. The OpenAI-compatible route is `/v1/models`, UAR uses `/api/uar/providers/{id}/models`, and BossFang uses `/api/models`. Native discovery describes configured availability, not successful inference.

diff --git a/docs/generated/runtime-reference.md b/docs/generated/runtime-reference.md
@@ -8,7 +8,7 @@ Release: **1.10.0**
 
 Generated by: `npm run docs:sync`
 
-Skill definitions: **153**
+Skill definitions: **194**
 
 Plugin targets: **14**
 

diff --git a/docs/guide/08-skills-overview.md b/docs/guide/08-skills-overview.md
@@ -1,6 +1,6 @@
 # 08 · Skills Overview
 
-A skill is a self-contained, portable unit of capability: a `SKILL.md` manifest with YAML frontmatter, optionally accompanied by `scripts/`, `references/`, and `assets/`. The prometheus-skill-pack ships 35 top-level skills across 13 categories — 95+ skills counting sub-skills — and every one of them is documented in the three pages that follow this one. This page explains the model: how skills are structured, how they are discovered, the standard they conform to, and where to find each category.
+A skill is a self-contained, portable unit of capability: a `SKILL.md` manifest with YAML frontmatter, optionally accompanied by `scripts/`, `references/`, and `assets/`. The generated full distribution index currently lists 215 skills. The UI/UX routing catalog contributes 41 entries, with a 40-entry portable subset shared with mini; its native Impeccable integration is full-only. Use the generated catalogs for current inventories and the [UI/UX guide](25-ui-ux-routing.md) for selective loading. This page explains the model: how skills are structured, how they are discovered, the standard they conform to, and where to find each category.
 
 ## The skills model
 
@@ -80,6 +80,7 @@ Every native skill conforms strictly to the [AgentSkills.io specification](https
 | **Research** | 1 | deep-research (10 sub-skills) | [10](10-language-skills.md) |
 | **Documentation** | 1 | llm-wiki | [10](10-language-skills.md) |
 | **Flint** | 6 | flint-sdk-csharp, -dart, -go, -kotlin, -swift, -ts | [10](10-language-skills.md) |
+| **UI/UX routing catalog** | 41 | Context, Pro Max, craft, taste, platform and review guidance; 40 portable entries | [25 · UI/UX routing](25-ui-ux-routing.md) |
 | **Imported (submodules)** | 2 | artifact-refiner, sycophancy-correction | [11](11-artifact-refiner.md), [07](07-sycophancy-correction.md) |
 
 ## Native, imported, and forge-rs skills

diff --git a/docs/guide/24-agent-teams.md b/docs/guide/24-agent-teams.md
@@ -11,7 +11,7 @@ The four procedures are available as ordinary skills through the process plugin
 
 | Procedure | Use it to |
 | --- | --- |
-| `agent-team-creator` | Describe an outcome, edit the proposed roles and stage native definitions. |
+| `agent-team-creator` | Describe an outcome, stage native definitions, then install or adopt the project team. |
 | `agent-team-manage` | Assign and track revisioned tasks, dependencies and evidence. |
 | `agent-team-models` | Compare configured models against explicit capabilities, tiers and price ceilings. |
 | `agent-team-handoff` | Prepare fresh context and transfer ownership after destination acceptance. |
@@ -52,6 +52,24 @@ Targets are UAR, Codex, Claude Code, Copilot, Kimi Code, MiniMax, OpenCode, Deep
 
 Native role overrides, team options and opaque files preserve settings beyond the common manifest, with source/version provenance. Preservation does not certify a setting against an installed tool. The [native contract reference](https://github.com/Prometheus-AGS/prometheus-skill-system/blob/main/skills/process/agent-team-creator/references/native-harnesses.md) links official sources and explains which options become proposed config files and which remain sidecar data for deliberate application.
 
+## Install or adopt a project team
+
+After reviewing the export, finish normal project-team creation with `install-project`. For a new team, save `{"project":"/path/to/project","team":<manifest>}` as `install-request.json`, then run:
+
+```text
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --input install-request.json --dry-run
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --input install-request.json
+node skills/process/agent-team-creator/scripts/cli.mjs install-project --project "/path/to/project" --check
+```
+
+For an existing team, use `--project "/path/to/project"` without an input manifest. A recorded `.agent-team/project-routing.json` selection wins; otherwise a sole `.agent-team/<id>/team.json` is adopted. Multiple candidates require `--team <id>`; stale selections fail rather than silently switching. Intentional manifest replacement requires `updateTeam: true` in the request. Creator check exits 2 for drift and 1 for errors.
+
+Installation writes managed discovery pointers to both instruction entrypoints, the active routing record and missing native definitions. Existing native files, role IDs, ownership, model policies, permissions and concurrency remain intact; differing configuration is reported for deliberate merge. Recovery records retain prior instruction bytes. Export stages proposals; installation establishes discovery. Neither starts execution or activates UAR/BossFang registration.
+
+All code tasks use the selected team’s relevant roles. UI roles conditionally use `prometheus-ui-ux`; UI review uses `prometheus-ui-review` after the whole implementation phase, without taste or user-only skill preloads. Backend work loads no UI guidance. Use native delegation only if available, otherwise disclose sequential role execution; builder self-review is not independent review. Zed receives the pointer in its effective existing instruction file. External ACP agents keep native configuration, and Zed parallel threads are not a delegation API.
+
+See [UI/UX routing](25-ui-ux-routing.md) for the selective workflow.
+
 ## Models, ownership and continuity
 
 Model selection combines team, role, skill and task policy. Strength tiers are explicit annotations; they are not inferred from a model's name. Unknown prices cannot satisfy a cost ceiling, and unavailable capability metadata remains unknown. Discovery is not a successful inference test. Review the selected exact model ID before updating policy and exporting it.