tech-lead-stack 1.0.1

package/.agents/workflows/reflexion-loop-sub-max.md

@@ -0,0 +1,46 @@
+---
+name: reflexion-loop-sub-max
+description: >-
+  [LOOP · SUB-MAX · NO API KEYS · CROSS-MODEL VERIFY] $100/mo tier
+  context-isolated plan hardening loop with multi-vendor cross-model
+  verification.
+modes:
+  - write
+---
+
+// turbo
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/reflexion-loop-sub-max.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition (FIRST ACTION — NO EXCEPTIONS)**:
+   Call the `get_skills` / `get_skill` tool with all 4 required fields:
+   - `skillName`: "reflexion-loop-sub-max"
+   - `projectName`: "<YOUR_CURRENT_PROJECT_NAME>"
+   - `model`: "<YOUR_MODEL_NAME>"
+   - `agent`: "<YOUR_AGENT_NAME>"
+
+2. **Pre-Flight Model Contract**: Formulate and print generator and critic model assignments and isolation levels (L0-L3). Check `CLAUDE_CODE_SUBAGENT_MODEL` (warn and cap at L2 if overridden). Resolve the critic with `./.ai/rtk-run run resolve-critic --writer <vendor> --writer-model <generator-model>` (Gemini `gemini-cli`/`agy` -> `codex` -> Claude sub-agent/`claude-cli` -> another model -> same model + STRONG `criticAdvisory`); record `criticResolution` in `state.json`. Never ask for `GOOGLE_CLOUD_PROJECT` — for personal Google logins the probe uses `agy`.
+3. **Phase 0B: Discovery**: Run scoped discovery, write Findings Ledger (`.loop-out/<runId>/findings.md`), and establish stack diagnosis.
+4. **Phase 1-3: Generate, Critique & Adjudicate**: Execute plan generation and critic passes. Track models in state file (`.loop-out/<runId>/state.json`). Handle exhaustion Modes A/B/C via fallback ladder; mark degraded critic pass as PROVISIONAL.
+5. **Mandatory Disclosure Verdict**: Emit disclosure status (`Model separation held` vs `MODEL SEPARATION LOST`) as the VERY FIRST line of `.loop-out/<runId>/verdict.md` and chat output, followed by Provenance Table.
+6. **Adjudicate (Human-in-the-Loop)**: When run is complete, report final verdict.
+   > "The Reflexion loop concluded at revision **N** with score **S/10** (Status: **<PASSED|PROVISIONAL>**). Approve to proceed, or override the last fix and run another loop?"
+7. **On approval**, hand `.loop-out/<runId>/plan.md` to `planning-expert` or `vertical-slice-decomposer` to execute atomic task list.
+
+---
+
+**IDE Notes:**
+- **Antigravity**: Paste into Agent Manager > Customizations > Workflows and invoke with `/reflexion-loop-sub-max`.
+- **Cursor**: Auto-symlinked to `~/.cursor/skills/workflow-reflexion-loop-sub-max/`, invoke with `@workflow-reflexion-loop-sub-max`.
+- **Continue**: Auto-symlinked to `~/.continue/prompts/reflexion-loop-sub-max.prompt`, invoke from the prompt menu.

package/.agents/workflows/reflexion-loop-sub-pro.md

@@ -0,0 +1,45 @@
+---
+name: reflexion-loop-sub-pro
+description: >-
+  [LOOP · SUB-PRO · NO API KEYS · CROSS-MODEL VERIFY] $20/mo tier
+  context-isolated loop with single-pass cross-model plan check.
+modes:
+  - write
+---
+
+// turbo
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/reflexion-loop-sub-pro.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition (FIRST ACTION — NO EXCEPTIONS)**:
+   Call the `get_skills` / `get_skill` tool with all 4 required fields:
+   - `skillName`: "reflexion-loop-sub-pro"
+   - `projectName`: "<YOUR_CURRENT_PROJECT_NAME>"
+   - `model`: "<YOUR_MODEL_NAME>"
+   - `agent`: "<YOUR_AGENT_NAME>"
+
+2. **Pre-Flight Model Contract**: Print model assignments and isolation levels (L0-L3). Check `CLAUDE_CODE_SUBAGENT_MODEL` (warn and cap at L2 if overridden). Resolve the critic with `./.ai/rtk-run run resolve-critic --writer <vendor> --writer-model <generator-model>` (Gemini `gemini-cli`/`agy` -> `codex` -> Claude sub-agent/`claude-cli` -> another model -> same model + STRONG `criticAdvisory`); record `criticResolution` in `.loop-out/<runId>/state.json`.
+3. **Phase 0B: Discovery**: Run scoped discovery, write Findings Ledger to `.loop-out/<runId>/loop.md`, and establish stack diagnosis.
+4. **Phase 1-3: Generate & Single-Pass Critique**: Execute single-pass generator and critic. On Mode B quota limit, consolidate into `.loop-out/<runId>/loop.md`, mark status `PARKED`, and report reset window.
+5. **Mandatory Disclosure Output**: Emit disclosure status (`Model separation held` vs `MODEL SEPARATION LOST`) as the VERY FIRST line of `.loop-out/<runId>/loop.md` and chat output, followed by Provenance Table.
+6. **Adjudicate (Human-in-the-Loop)**: Report final status to human.
+   > "The Reflexion loop concluded with score **S/10** (Status: **<PASSED|PROVISIONAL>**). Approve to proceed, or override and run another pass?"
+7. **On approval**, hand `.loop-out/<runId>/loop.md` to `planning-expert` or `vertical-slice-decomposer` to execute atomic task list.
+
+---
+
+**IDE Notes:**
+- **Antigravity**: Paste into Agent Manager > Customizations > Workflows and invoke with `/reflexion-loop-sub-pro`.
+- **Cursor**: Auto-symlinked to `~/.cursor/skills/workflow-reflexion-loop-sub-pro/`, invoke with `@workflow-reflexion-loop-sub-pro`.
+- **Continue**: Auto-symlinked to `~/.continue/prompts/reflexion-loop-sub-pro.prompt`, invoke from the prompt menu.

package/.agents/workflows/reflexion-loop.md

@@ -0,0 +1,66 @@
+---
+name: reflexion-loop
+description: >-
+  [LOOP · DUAL-MODEL · API KEYS] ✨ Special feature — run the two-model
+  self-correcting plan loop (Gemini drafts, Claude grades). Requires
+  GEMINI_API_KEY + ANTHROPIC_API_KEY.
+modes:
+  - write
+---
+
+// turbo
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/reflexion-loop.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition (FIRST ACTION — NO EXCEPTIONS)**:
+   Call the `get_skills` / `get_skill` tool with all 4 required fields:
+   - `skillName`: "reflexion-loop"
+   - `projectName`: "<YOUR_CURRENT_PROJECT_NAME>"
+   - `model`: "<YOUR_MODEL_NAME>"
+   - `agent`: "<YOUR_AGENT_NAME>"
+
+2. **Phase 0B: Confirm the brief.** Take the user's feature request / ticket as
+   the brief. If they pointed at a file, read it.
+
+3. **Run the engine (terminal).** From the repo root:
+
+   ```
+   rtk run reflexion-loop -- "<BRIEF TEXT>"
+   ```
+
+   (Equivalent: `npx tsx scripts/reflexion-loop.ts "<BRIEF TEXT>" --repo . --max 3 --threshold 8`.)
+   The engine reads the repo for Phase-0 diagnosis, then loops Gemini (writer) ↔
+   Claude (critic) until it passes or caps, and Claude writes the final verdict.
+
+4. **Surface the artifacts** from `.reflexion-out/`:
+   - `plan.md` — the final implementation plan.
+   - `diminishing-returns.svg` — score-per-revision curve.
+   - The adjudicator verdict printed at the end of the run.
+
+5. **Adjudicate (Human-in-the-Loop).** Report:
+   > "The Reflexion loop concluded at revision **N** with score **S/10**. Approve
+   > to proceed, or override the last fix and run another loop?"
+   Handle the run's exit state correctly:
+   - **0 (Passed/user-approve):** The plan is ready.
+   - **2 (Parked/awaiting answers):** The adjudicator has questions. Instruct the user to edit `interview.md` and resume.
+   - **3 (Budget-exceeded/user-stop):** The budget cap was tripped or manually stopped. Summarise the gap.
+   - **4 (Refine-contract-violation/internal error):** The loop failed strict verification.
+
+5. **On approval**, hand `.reflexion-out/plan.md` to `planning-expert` or
+   `vertical-slice-decomposer` to execute the atomic task list.
+
+> [!NOTE] **This is the developer path.** Running here (IDE + MCP) means the
+> agent may go on to **change code** from the reviewed plan, and the run is
+> **logged to Prisma** (`source: 'mcp'`). The website/chat version is read-only
+> and only ever returns a plan + an IDE prompt — it never edits code.

package/.agents/workflows/regression-bug-fix.md

@@ -0,0 +1,31 @@
+---
+name: regression-bug-fix
+description: Unified Feedback & Regression Fix
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/regression-bug-fix.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "regression-bug-fix"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 1: Environment Discovery**: Identify the tech stack by reading root configuration files (e.g., package.json, pyproject.toml, go.mod, Cargo.toml, pom.xml, build.gradle) to understand architectural constraints.
+
+3. Follow its unified workflow to resolve regressions with high-integrity verification.

package/.agents/workflows/security-audit.md

@@ -0,0 +1,31 @@
+---
+name: security-audit
+description: Security Audit
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/security-audit.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "security-audit"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 1: Environment Discovery**: Identify the tech stack by reading root configuration files (e.g., package.json, pyproject.toml, go.mod, Cargo.toml, pom.xml, build.gradle) to understand architectural constraints.
+
+3. Follow its workflow to scan for malicious patterns and potential ecosystem vulnerabilities.

package/.agents/workflows/standup-daily-summary.md

@@ -0,0 +1,31 @@
+---
+name: standup-daily-summary
+description: Daily Standup Report
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/standup-daily-summary.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "daily-standup"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 1: Environment Discovery**: Identify the tech stack by reading root configuration files (e.g., package.json, pyproject.toml, go.mod, Cargo.toml, pom.xml, build.gradle) to understand architectural constraints.
+
+3. Follow its workflow to generate a rolling 2-day standup report from git activity.

package/.agents/workflows/strategy-target-evaluation.md

@@ -0,0 +1,31 @@
+---
+name: strategy-target-evaluation
+description: Product Strategy Audit
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/strategy-target-evaluation.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "product-strategist"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 1: Environment Discovery**: Identify the tech stack by reading root configuration files (e.g., package.json, pyproject.toml, go.mod, Cargo.toml, pom.xml, build.gradle) to understand architectural constraints.
+
+3. Follow its workflow to validate market positioning and roadmap prioritization.

package/.agents/workflows/style-logic-exporter.md

@@ -0,0 +1,84 @@
+---
+name: style-logic-exporter
+description: Export Tailwind v3.4 design tokens to Figma (Tokens Studio format)
+modes:
+  - write
+---
+
+# 🎨 Figma Token Exporter
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/style-logic-exporter.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "style-logic-exporter"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+> Supersedes the previous "Style Logic Exporter" skill. This workflow produces a
+> machine-readable `figma-tokens.json` in **Tokens Studio v2** format, directly
+> importable into Figma via the [Tokens Studio plugin](https://tokens.studio/).
+> Compatible with Tailwind v3.4.
+
+## Objective
+To prevent design decay — when Tailwind tokens in code silently diverge from what
+designers see in Figma — by providing an automated, repeatable export of the project's
+current token state directly from `tailwind.config.ts`.
+
+---
+
+## ⛔️ Guardrails
+1. Always export from the **target client project**, not from the Tech-Lead Stack itself.
+2. The `figma-tokens.json` output is read-only for designers. Designers should
+   **not** edit the JSON manually; changes must originate in `tailwind.config.ts`.
+3. After exporting, run `pnpm tsc --noEmit` on the target project to confirm no
+   config regressions were introduced.
+
+---
+
+## 🛠 Execution Steps
+
+### Step 1: Identify the Target Project
+- Confirm the absolute path to the target project (e.g., `/Users/dev/repos/gilly/client`).
+- Verify a `tailwind.config.ts` (or `.js`) exists at the root.
+
+### Step 2: Run the Token Exporter
+From the **Tech-Lead Stack** root directory, execute:
+```bash
+npx ts-node scripts/export-tokens.ts --project <ABSOLUTE_PATH_TO_TARGET_PROJECT>
+```
+*(Example: `npx ts-node scripts/export-tokens.ts --project /Users/dev/repos/gilly/client`)*
+
+### Step 3: Verify the Output
+- Confirm `figma-tokens.json` was created in the target project root.
+- Check the console output for the token count — a well-configured Tailwind project
+  should export 50+ tokens.
+- Open the file and spot-check that color values match the known brand palette.
+
+### Step 4: Provide the JSON to the Designer
+Report the output to the designer/PM with the following message:
+> *"Tokens exported successfully. Import `figma-tokens.json` from the repo root
+> into Figma via: Tokens Studio Plugin → `Load` → `From file`. The file contains
+> [N] tokens covering colors, spacing, typography, border radius, and font families."*
+
+### Step 5: Chromatic Sync Note (Optional)
+If this export is being done as part of a design review, remind the team:
+> *"After importing to Figma, run a Chromatic snapshot to confirm the rendered
+> component matches the updated token values."*
+
+---
+**Completion:** Summarize what was exported (token count breakdown by category) and
+confirm the `figma-tokens.json` file path.

package/.agents/workflows/ui-spec-generator.md

@@ -0,0 +1,156 @@
+---
+name: ui-spec-generator
+description: AI-Powered UI Spec Generator
+modes:
+  - write
+---
+
+# 🚀 AI-Powered UI Spec Generator
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/ui-spec-generator.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "ui-spec-generator"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+**SECURITY CLEARANCE: PM, DESIGNER, OR DEVELOPER INITIATION**
+This is the **only** workflow permitted to edit code directly from the Tech-Lead Stack Agent Chat. It must be executed with extreme caution and adhere strictly to branching and styling protocols.
+
+## Objective
+To take natural language feature requirements from a PM and generate a "Technical UI Spec" consisting of functional, base-theme Shadcn/Radix components placed in the **correct location** for the target project's design system. This allows developers to immediately begin wiring up state and logic without waiting for pixel-perfect Figma designs.
+
+## ⛔️ STRICT GUARDRAILS
+1. **NEVER** edit code on the `main` branch.
+2. **ALWAYS** create a new branch before modifying any files.
+3. **ONLY** use base Shadcn primitives and the project's brand tokens (Tailwind classes). Do not introduce arbitrary custom CSS or third-party component libraries without permission.
+4. **ALWAYS** report the resolved component output path to the PM and wait for confirmation before creating any files.
+
+---
+
+## 🛠 Execution Steps
+
+### Step 1: Pre-Flight & Discovery
+
+#### 1a. Base Constraints
+Review the following files to understand the target project's architecture:
+- `package.json` — detect package manager (`pnpm`, `npm`, `yarn`), workspaces config, and monorepo structure.
+- `components.json` — Shadcn configuration and component alias paths.
+- `tailwind.config.ts` — brand tokens and theme extensions.
+
+#### 1b. Component & Design Mapping
+- Analyze the requirements and map them to standard Shadcn components.
+- **Source of Truth Check**: If a Figma Link was provided during initiation, scrape the link (using `firecrawl_scrape`) to extract layout, colors, and typography constraints before generating code.
+
+
+#### 1c. 🗂️ Design System Discovery (3-Tier Protocol)
+Determine where new UI atom components should be created. Execute the following tiers **in order**, stopping at the first match:
+
+**Tier 1 — Project Settings (Highest Priority)**
+Query the Tech-Lead Stack API for the project's configured `designSystemPath`:
+```bash
+curl -s <TECH_LEAD_STACK_URL>/api/projects/<PROJECT_ID>/settings \
+  -H "Authorization: Bearer $TECH_LEAD_API_KEY"
+```
+If `settings.designSystemPath` is set (e.g. `libs/gilly-ui/src/components`), **use it directly**. Skip Tier 2 and Tier 3.
+
+**Tier 2 — Monorepo Auto-Detection**
+If no configured path, inspect the project root for monorepo signals:
+- `pnpm-workspace.yaml` or `package.json` `"workspaces"` field
+- `turbo.json` or `nx.json`
+- A `packages/` or `libs/` directory
+
+If a monorepo is detected, find the UI library package by looking for:
+- A sub-package whose `package.json` `"scripts"` includes `storybook`
+- A sub-package with a `src/components/` directory and UI-related dependencies (e.g., `@radix-ui/*`, `class-variance-authority`)
+
+Use `<detected-ui-lib>/src/components/<feature>/` as the target path.
+
+**Tier 3 — Fallback (Single App)**
+If no monorepo is detected:
+- Check if `src/components/` exists → use `src/components/<feature>/`
+- Check if `components/` exists → use `components/<feature>/`
+- If neither exists → **create** `src/components/<feature>/` and use it.
+
+#### 1d. Confirm Output Path With PM
+Before proceeding to branching, report the resolved path to the PM:
+> *"I have resolved the component output path to: `<resolved-path>`. Is this correct?"*
+
+**Do not proceed to Step 2 until the PM confirms.**
+
+#### 1e. Codebase Pattern Review
+Review 2–3 existing components in the resolved output path to understand:
+- File naming conventions (`kebab-case.tsx` vs `PascalCase.tsx`)
+- Import patterns (barrel exports via `index.ts` vs direct imports)
+- Whether Storybook story files are co-located (`ComponentName.stories.tsx`)
+
+---
+
+### Step 2: Branching Protocol
+You **must** execute these commands to ensure a clean workspace before editing code:
+```bash
+git checkout main
+git pull
+git checkout -b feature/ui-spec-<feature-name>
+```
+*(Replace `<feature-name>` with a concise, kebab-case representation of the feature).*
+
+---
+
+### Step 3: Skeleton Generation
+1. For each required Shadcn component, run the add command using the project's detected package manager **before** writing any code that imports it:
+   ```bash
+   npx shadcn-ui@latest add <component>
+   ```
+2. Implement the requested feature inside the resolved output path from Step 1c.
+   - Place **page-level** files at: `app/(routes)/[feature]/page.tsx`
+   - Place **reusable UI atoms** at: `<resolved-component-path>/<feature>/`
+   - If Storybook stories are co-located in this project, create a `<ComponentName>.stories.tsx` stub alongside each new atom.
+3. Use dummy data or mocked interfaces where backend APIs are not yet defined.
+4. Ensure all code is strictly typed (TypeScript) and uses Tailwind CSS for layout.
+5. **Figma Alignment**: If a Figma link was provided, ensure the generated Tailwind classes match the extracted tokens from Step 1b.
+6. Follow the exact naming and import patterns identified in Step 1e.
+
+---
+
+### Step 4: Commit & Push
+Once the code is implemented and verified locally via linting or visual inspection (if applicable):
+```bash
+git add .
+git commit -m "feat(ui-spec): generate base skeleton for <feature-name>"
+git push -u origin feature/ui-spec-<feature-name>
+```
+
+---
+
+### Step 5: Task Handoff & Dev Notification
+1. Generate a `task.md` document for the developer in the root of the target project outlining:
+   - **Component Location:** `<resolved-component-path>/<feature>/`
+   - **Note:** All Shadcn/Radix components have already been added and installed. The developer only needs to run `pnpm install` (or the project's package manager) to sync their local `node_modules`.
+   - The data wiring required to make the skeleton fully functional (API calls, Prisma queries, React Query hooks, form actions, etc.).
+2. Hit the Tech-Lead Stack API to notify the developers that the branch is ready:
+   ```bash
+   curl -X POST <TECH_LEAD_STACK_URL>/api/notify/devs \
+     -H "Content-Type: application/json" \
+     -d '{"projectId": "<YOUR_PROJECT_ID>", "branchName": "feature/ui-spec-<feature-name>", "message": "UI Spec generated for <feature-name>. Components at: <resolved-component-path>/<feature>/"}'
+   ```
+
+---
+**Completion:** Summarize the operation for the PM in the chat, including:
+- The branch name
+- The exact component output path used (and which discovery tier resolved it)
+- Confirmation that the developers have been notified

package/.agents/workflows/verify-changes.md

@@ -0,0 +1,31 @@
+---
+name: verify-changes
+description: Visual Smoke Test
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/verify-changes.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "visual-verifier"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 1: Environment Discovery**: Identify the tech stack by reading root configuration files (e.g., package.json, pyproject.toml, go.mod, Cargo.toml, pom.xml, build.gradle) to understand architectural constraints.
+
+3. Follow its workflow to capture before/after screenshots and upload evidence of ecosystem parity.

package/.agents/workflows/vertical-slice.md

@@ -0,0 +1,52 @@
+---
+name: vertical-slice
+description: Decompose user stories into ClickUp-ready vertical slices
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/vertical-slice.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+**Precondition:** Requires a formalized spec. If you receive a raw story, run the `specify` phase (e.g., `solutioning-facilitator`) first.
+**Multi-Repo Grounding:** Expect and enumerate multiple workspace roots; do not guess across unmounted repositories.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` / `get_skill` tool:
+   - skillName: "vertical-slice-decomposer"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+2. **Phase 0 (cont): Domain & Infra Discovery**: Identify the tech stack, the
+   team's domain boundary (UI contract vs API contract), the dark-release gate
+   (`beta_*` cookies / `x-beta-flags`), and the mocking layer (backend-first vs
+   MSW fallback). If the user attaches **design screenshots or Figma URLs**,
+   treat them as in-scope spec: read Figma via the Figma MCP/connector when
+   available (else request an exported frame), and extract the states/variants
+   shown — each is a candidate slice boundary.
+
+3. **Run the skill workflow**: Build the Slice Ledger from the input story/
+   stories and any design frames, slice vertically (deployability test + BDD
+   boundaries + design states), decide beta-flag and mock-vs-real per slice, and
+   emit ClickUp-ready task blocks per the Output Contract (including the Design
+   reference field).
+
+4. **Anti-drift**: This is iterative. After any detour into a single slice's
+   detail or error resolution, fold the result back into that slice, reprint the
+   Ledger, and resume the decomposition queue. Never abandon a pending slice.
+
+5. **Deliver**: A `vertical-slices.md` handoff (paste-ready for ClickUp) plus the
+   final Ledger. Do NOT auto-create ClickUp items unless a ClickUp connector is
+   present AND the user explicitly confirms.

package/.agents/workflows/weekly-leadership-report.md

@@ -0,0 +1,39 @@
+---
+name: weekly-leadership-report
+description: Weekly Leadership Status Report (Team-Wide)
+modes:
+  - write
+---
+
+// turbo-all
+
+**CRITICAL: PHASE 0 - SKILL ACQUISITION IS NON-NEGOTIABLE.**
+**YOU MUST CALL THE GET_SKILLS TOOL EVEN IF YOU ALREADY HAVE THE CONTEXT. FAILURE TO DO SO BYPASSES MISSION TELEMETRY.**
+
+> [!IMPORTANT]
+> **ANTI-CONFLATION DIRECTIVE:**
+> This file (`.agents/workflows/weekly-leadership-report.md`) is a workflow launcher stub, NOT the skill definition.
+> Viewing this file via `view_file` does NOT satisfy Phase 0. You MUST call the MCP `get_skills` or `get_skill` tool FIRST with all 4 required fields before executing any other steps.
+
+- **Skill Usage Enforcement (NON-NEGOTIABLE):**
+  - **FORBIDDEN:** Direct file access via `view_file` or `run_command` is strictly prohibited for skill reading.
+  - **IDE / MCP-enabled Agent:** You MUST call the MCP `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing).
+  - **Chat UI (/chat):** You MUST call the internal `get_skill` tool.
+
+1. **Phase 0: Skill Acquisition**: Call the `get_skills` tool (which may be prefixed as `mcp_tech-lead-stack_get_skills` or `tech-lead-stack_get_skills` depending on client prefixing):
+   - skillName: "weekly-leadership-report"
+   - projectName: "<YOUR_CURRENT_PROJECT_NAME>"
+   - model: "<YOUR_MODEL_NAME>"
+   - agent: "<YOUR_AGENT_NAME>"
+
+3. **Phase 2: Data Extraction (Holistic)**:
+   - **DO NOT** output any plans and **DO NOT** use subagents. You must directly gather the data yourself.
+   - Use the `mcp_chrome-devtools-mcp_navigate_page` and `mcp_chrome-devtools-mcp_take_screenshot` (or related MCP tools) to view the provided ClickUp URLs using the user's existing authenticated browser session.
+   - Run `git log origin/main --since="7 days ago" --oneline` to see the entire team's merged work.
+   - Run `git tag --sort=-creatordate` to identify version spans.
+
+4. **Phase 3: Synthesis**:
+   - Follow the instructions in the `weekly-leadership-report` skill.
+   - Categorize work into DEVS, QA, and DESIGN sections.
+   - **DO NOT** output any planning steps or subagent tasks. Only output the final result.
+   - **FORCE OUTPUT INTO A MARKDOWN CODE BLOCK** for easy copy-pasting.