Integrate PraxStack skills-and-personas + tiered agent skill stack

.agents/skills/.gitignore

@@ -0,0 +1,48 @@
+# PraxStack install-time skills (generated by scripts/install-praxstack-skills.sh).
+# Not vendored in git — refresh via ./scripts/install-praxstack-skills.sh or extended tier.
+# When praxstack-skills.lock.json adds skills, update this file in the same commit.
+/apex-autonomous-mode/
+/autonomous-orchestrion/
+/backend-architecture-standards/
+/backend-pe-cpp/
+/backend-pe-java/
+/backend-pe-javascript/
+/backend-pe-nodejs/
+/backend-pe-python-ml/
+/backend-pe-python/
+/backend-pe-typescript/
+/backend-pe/
+/backend-system-design-expert/
+/baron-von-markup/
+/blueprint-creator/
+/chronicle/
+/coding-agent-leadership-principles/
+/concept-cartographer/
+/constellation-team/
+/cross-agent-handoff/
+/devops-sre-engineer/
+/frontend-design-excellence/
+/frontend-excellence-standards/
+/frontend-pe/
+/frontend-uiux-designer/
+/gabriel-petersson-topdown-mentor/
+/idea-capturer/
+/kingmode/
+/lecture-alchemist/
+/mental-health-screening-companion/
+/obsidian-cli/
+/orchestrion-universal-agent-router/
+/principal-engineer/
+/product-manager/
+/professor-alex-interview/
+/qa-security-engineer/
+/security-compliance-standards/
+/spec-creator/
+/super-mode-core/
+/superimprove/
+/svg-logo-designer/
+/teach-pro-max/
+/techtutor/
+/transcribe-refiner/
+/transcript-pipeline/
+/ultrathink-frontend/

.agents/skills/README.md

@@ -39,15 +39,55 @@ discover → interrogate/spec → plan → implement → review → security →
 
 Also installed: **Anthropic dev subset** (MCP builder, frontend-design, webapp-testing, etc.) and **awesome-copilot** (417-skill toolbox).
 
+## Install tiers
+
+The install script supports three tiers via `BERD_SKILLS_TIER`:
+
+| Tier | Env value | What installs |
+| --- | --- | --- |
+| **CORE** (default) | `core` | Layered pipeline packs (~740 skills) |
+| **EXTENDED** | `extended` | CORE + selective S-tier additions below |
+| **ALL** | `all` | EXTENDED + optional hooks (no bulk verticals) |
+| **PRAXSTACK** | `praxstack` | PraxStack skills-and-personas only (no core packs) |
+
+```bash
+./scripts/install-agent-skills.sh                    # CORE (default)
+BERD_SKILLS_TIER=extended ./scripts/install-agent-skills.sh
+BERD_SKILLS_TIER=praxstack ./scripts/install-agent-skills.sh   # PraxStack layer only
+BERD_PRAXSTACK_SKILLS=1 ./scripts/install-agent-skills.sh      # CORE + PraxStack
+./scripts/install-praxstack-skills.sh                          # PraxStack only (direct)
+```
+
+The script pins `npx skills` to the version in `skills-lock.json` (`skillsCli` field). PraxStack content is pinned in `praxstack-skills.lock.json`. Override with `SKILLS_CLI_VERSION=latest` if needed.
+
+## Extended tier additions
+
+| Skill | Source | Role |
+| --- | --- | --- |
+| **hallmark** | [nutlope/hallmark](https://github.com/nutlope/hallmark) | UI taste gate — anti-AI-slop design for greenfield pages and redesigns |
+| **deep-research** | [24601/agent-deep-research](https://github.com/24601/agent-deep-research) | Async deep research (one skill, not four parallel systems) |
+| **last30days** | [mvanhorn/last30days-skill](https://github.com/mvanhorn/last30days-skill) | Recency radar — what people say about a topic in the last 30 days |
+
+## Conflict warnings
+
+Do **not** run multiple full methodologies on the same task — pick one primary workflow:
+
+| Task type | Pick one | Avoid stacking |
+| --- | --- | --- |
+| Planning | pstack `/poteto-mode` **or** superpowers `/writing-plans` **or** gstack `/plan-ceo-review` | All three in one session |
+| Code review | Berd `code-review` **or** compound `ce-code-review` **or** mattpocock `/code-review` | Redundant review passes |
+| UI design | hallmark **or** shadcn `/improve` design sections | Duplicate design audits |
+| Research | `deep-research` **or** `last30days` **or** compound research agents | Four parallel research systems |
+
 ## Install or refresh
 
 ```bash
 ./scripts/install-agent-skills.sh
 ```
 
-The script uses [`npx skills@latest`](https://github.com/vercel-labs/skills) with `-a cursor -y --copy` (Cursor only, project-local). It restores Berd-owned skills after upstream packs are copied and removes junk agent directories.
+The script uses [`npx skills`](https://github.com/vercel-labs/skills) (pinned via `skills-lock.json`) with `-a cursor -y --copy` (Cursor only, project-local). It restores Berd-owned skills after upstream packs are copied and removes junk agent directories.
 
-Run twice to verify idempotence. Expect ~740 skills after a full install.
+Run twice to verify idempotence. Expect ~740 skills after CORE; ~743 after EXTENDED.
 
 ## One-time repo setup
 
@@ -84,6 +124,24 @@ Re-run those skills if you switch issue trackers or change pstack model assignme
 - `create-pr` — open and watch PRs
 - `experimental-features` — experiment registry workflow
 
+## PraxStack layer (skills-and-personas)
+
+[praxstack/skills-and-personas](https://github.com/praxstack/skills-and-personas) is the canonical source for PraxStack workflows: orchestration (`kingmode`, `constellation-team`), role skills (`principal-engineer`, `product-manager`, …), learning (`teach-pro-max`, `techtutor`), and leadership (`superimprove`, `cross-agent-handoff`).
+
+| Artifact | Install location | Install script |
+| --- | --- | --- |
+| **Skills** (~42 canonical + 4 public) | `.agents/skills/` | `scripts/install-praxstack-skills.sh` |
+| **Personas** (multi-file packs) | `docs/agents/personas/praxstack/` | same |
+| **md-personas** (single-file) | `docs/agents/personas/md/` | same |
+| **Goals / workflows** (alignment, high-end-operator) | `docs/agents/workflows/praxstack/` | same |
+| **Goals guide** | `docs/agents/goals.md` | checked in |
+
+Pin: `praxstack-skills.lock.json`. Bump with `./scripts/update-praxstack-skills-lock.sh <ref>`.
+
+Installed by default when `BERD_SKILLS_TIER=extended|all|praxstack`, or force with `BERD_PRAXSTACK_SKILLS=1`. Cloud Agent `cloud-agent-install.sh` runs the PraxStack installer.
+
+**Conflict note:** PraxStack `kingmode` / `super-mode-core` overlap conceptually with pstack `/poteto-mode` and superpowers `/writing-plans` — pick one primary planner per task (see conflict table above).
+
 ## agent-browser CLI
 
 The browser QA layer needs the CLI alongside the skill:
@@ -156,13 +214,40 @@ Install manually when relevant:
 npx skills@latest add supabase/agent-skills --skill '*' -a cursor -y --copy
 npx skills@latest add cloudflare/skills --skill '*' -a cursor -y --copy
 npx skills@latest add aws/agent-toolkit-for-aws/skills --skill '*' -a cursor -y --copy
+# Remotion vertical (12 skills — install one, not all):
+npx skills@latest add remotion-dev/skills --skill remotion-best-practices -a cursor -y --copy
+# NVIDIA (343 skills — pick individual skills only):
+npx skills@latest add nvidia/skills --list
+npx skills@latest add nvidia/skills --skill <name> -a cursor -y --copy
 ```
 
+## MCP servers
+
+Project MCP config: `.cursor/mcp.json`
+
+| Server | Command | Purpose |
+| --- | --- | --- |
+| **Context7** | `npx -y @upstash/context7-mcp` | Up-to-date library documentation lookup |
+
+## Optional runtime tools (manual)
+
+These are **not** run by the install script. Install when needed:
+
+| Tool | Install | Notes |
+| --- | --- | --- |
+| **OpenSpec** | `npm install -g @fission-ai/openspec@latest` then `openspec init` | Creates `openspec/` structure for spec-driven workflows |
+| **Graphify** | `uv tool install graphifyy` then `graphify cursor install --project` | Requires `uv` on PATH; visualizes codebase graphs |
+| **Impeccable** | `npx impeccable skills install` (project) then `/impeccable init` in Agent chat | Design-system skill; installs to `.cursor/` and `.agents/` |
+| **agent-browser CLI** | `npm install -g agent-browser` | Required for browser QA layer (see above) |
+
 ## Searchable armory (do NOT bulk-install)
 
 | Pack | Source | Notes |
 | --- | --- | --- |
 | **Microsoft skills** | [microsoft/skills](https://github.com/microsoft/skills) | Large repo — context rot risk. Use `npx skills add microsoft/skills --list` and install individual skills only. |
+| **wshobson/agents** | [wshobson/agents](https://github.com/wshobson/agents) | 94 plugins — do NOT bulk install |
+| **NVIDIA skills** | [nvidia/skills](https://github.com/nvidia/skills) | 343 GPU/AI-Q skills — install individual skills only |
+| **everything-claude-code** | various | Redundant with gstack + superpowers + pstack + compound — skip |
 | Vercel skills CLI | [vercel-labs/skills](https://github.com/vercel-labs/skills) | Package manager for all skills |
 | Anthropic skills | [anthropics/skills](https://github.com/anthropics/skills) | Dev subset installed; browse for more |
 | Supabase | [supabase/agent-skills](https://github.com/supabase/agent-skills) | Postgres/Supabase workflows |

.cursor/mcp.json

@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "context7": {
+      "command": "npx",
+      "args": ["-y", "@upstash/context7-mcp"]
+    }
+  }
+}

.gitignore

@@ -28,7 +28,14 @@ Thumbs.db
 
 # Scratch / working files
 .scratch/
+.cache/
 .docker-cache/
+
+# PraxStack install outputs (generated by scripts/install-praxstack-skills.sh)
+docs/agents/praxstack-manifest.json
+docs/agents/personas/praxstack/
+docs/agents/personas/md/
+docs/agents/workflows/praxstack/
 .docker-home/
 avatar-manifest.json
 project-artifacts-manifest.json

AGENTS.md

@@ -126,7 +126,22 @@ local override.
 
 ## Agent skills
 
-Project skills live in `.agents/skills/`. Install or refresh with `./scripts/install-agent-skills.sh`. **gstack** (`/plan-ceo-review`, `/ship`, `/qa`, …) also needs its runtime: `./scripts/install-gstack-runtime.sh` (runs automatically at the end of the skill install script). See `.agents/skills/README.md`.
+Project skills live in `.agents/skills/`. Install or refresh:
+
+```bash
+./scripts/install-agent-skills.sh                      # CORE tier (default)
+BERD_SKILLS_TIER=extended ./scripts/install-agent-skills.sh   # + hallmark, deep-research, last30days, PraxStack
+BERD_SKILLS_TIER=praxstack ./scripts/install-agent-skills.sh  # PraxStack layer only
+./scripts/install-praxstack-skills.sh                   # PraxStack skills, personas, workflows
+```
+
+**gstack** (`/plan-ceo-review`, `/ship`, `/qa`, …) also needs its runtime: `./scripts/install-gstack-runtime.sh` (runs automatically at the end of the skill install script when CORE tier runs).
+
+- Stack tiers, conflict warnings, and optional tools: `.agents/skills/README.md`
+- PraxStack goals, personas, workflows: `docs/agents/goals.md`
+- MCP servers (Context7): `.cursor/mcp.json`
+- Skills CLI version pin: `skills-lock.json` → `skillsCli`
+- PraxStack repo pin: `praxstack-skills.lock.json`
 
 ### Issue tracker
 

docs/agents/goals.md

@@ -0,0 +1,62 @@
+# Agent goals and workflow alignment
+
+Berd agents should align on **what** to build before **how**. PraxStack workflow material from [skills-and-personas](https://github.com/praxstack/skills-and-personas) is vendored as prompts (not duplicate skill bodies).
+
+## Goals layer (when to use)
+
+| Situation | Entry point | Location |
+| --- | --- | --- |
+| Reconnect after a long break; verify shared understanding | **Align** | `docs/agents/workflows/praxstack/project-alignment/ALIGN-ONLY.md` |
+| New machine / fresh clone; install packs + align + QA report | **Align + install + QA** | `docs/agents/workflows/praxstack/project-alignment/ALIGN-INSTALL-QA.md` |
+| Install named packs only (gstack, Superpowers, Matt Pocock, OpenSpec, deep-research) | **Install skills** | `docs/agents/workflows/praxstack/project-alignment/INSTALL-SKILLS.md` |
+| Steady engineering loop (spec → plan → build → review → ship) | **High-end operator** | `docs/agents/workflows/praxstack/high-end-operator/README.md` |
+
+These prompts **invoke** installed skills by name. They do not replace Berd's core stack in `.agents/skills/README.md`.
+
+## Steady-state loop
+
+For day-to-day Berd work, prefer the layered pipeline in `.agents/skills/README.md`:
+
+```
+discover → interrogate/spec → plan → implement → review → security → browser QA → ship → learn
+```
+
+Use PraxStack **goals / alignment** prompts when:
+
+- You no longer trust that agent and human want the same thing
+- A new contributor or cloud agent needs project reconstruction
+- You need constellation-team role routing (`kingmode`, `constellation-team`, `product-manager`, etc.)
+
+## PraxStack skills (canonical for these workflows)
+
+Installed by `./scripts/install-praxstack-skills.sh` into `.agents/skills/`:
+
+| Category | Examples | Slash / trigger |
+| --- | --- | --- |
+| Orchestration | `kingmode`, `super-mode-core`, `ultrathink-frontend`, `constellation-team` | Mode keywords or skill name |
+| Engineering roles | `principal-engineer`, `backend-pe`, `frontend-pe`, `qa-security-engineer`, `devops-sre-engineer` | Skill name |
+| Document production | `blueprint-creator`, `spec-creator`, `transcript-pipeline` | Skill name |
+| Learning | `teach-pro-max`, `techtutor`, `lecture-alchemist`, `gabriel-petersson-topdown-mentor` | `/teach-pro-max`, etc. |
+| Leadership | `coding-agent-leadership-principles`, `cross-agent-handoff`, `superimprove` | Skill name |
+
+Full inventory: run the installer, then `ls .agents/skills/ | grep -E 'kingmode|constellation|teach-pro'`.
+
+## Personas (reference material)
+
+Persona packs live under `docs/agents/personas/praxstack/` (multi-file) and `docs/agents/personas/md/` (single-file). They are **source material** for skills — not Cursor rules. Do not copy persona bodies into `.cursor/rules/`; use the distilled skills in `.agents/skills/` instead.
+
+## Pin and refresh
+
+```bash
+./scripts/install-praxstack-skills.sh
+# or with the full Berd stack (extended tier includes PraxStack):
+BERD_SKILLS_TIER=extended ./scripts/install-agent-skills.sh
+```
+
+Commit pin: `praxstack-skills.lock.json` → `ref` field. Bump with:
+
+```bash
+./scripts/update-praxstack-skills-lock.sh <commit-sha-or-tag>
+```
+
+Manifest after install: `docs/agents/praxstack-manifest.json`.

praxstack-skills.lock.json

@@ -0,0 +1,15 @@
+{
+  "version": 1,
+  "repo": "https://github.com/praxstack/skills-and-personas.git",
+  "ref": "78a7ee914ae96a56e7e1ad5c27c4dca4b603dc24",
+  "canonicalSkillsDir": "new-skills",
+  "additionalSkills": [
+    "teach-pro-max",
+    "superimprove",
+    "coding-agent-leadership-principles",
+    "cross-agent-handoff"
+  ],
+  "personasDir": "personas",
+  "mdPersonasDir": "md-personas",
+  "workflows": ["prompts/project-alignment", "prompts/high-end-operator"]
+}

scripts/cloud-agent-install.sh

@@ -14,5 +14,8 @@ export CI="${CI:-true}"
 just _setup-dev-deps
 GOOSE_DEV_MODE=required GOOSE_BUILD_PROFILE=debug ./scripts/ensure-local-goose.sh
 
+# PraxStack skills-and-personas (project-local skills, personas, workflow prompts).
+"$repo_root/scripts/install-praxstack-skills.sh"
+
 # gstack skill markdown lives in .agents/skills/; runtime installs to ~/.claude/skills/gstack.
 "$repo_root/scripts/install-gstack-runtime.sh" -q