tribunal-kit 5.7.0 → 5.8.0

package/.agent/ARCHITECTURE.md

@@ -42,7 +42,7 @@ Type any of these in your AI IDE chat:
 | ----------------------- | -------------------------------------------------------------------------------------- | ----------------------------------- |
 | `/generate`             | Full Tribunal: Maker → Parallel Review → Human Gate                                    | `workflows/generate.md`             |
 | `/review`               | Audit existing code (no generation)                                                    | `workflows/review.md`               |
-| `/tribunal-full`        | ALL 18 reviewers at once — maximum coverage                                            | `workflows/tribunal-full.md`        |
+| `/tribunal-full`        | ALL 19 reviewers at once — maximum coverage                                            | `workflows/tribunal-full.md`        |
 | `/tribunal-backend`     | Logic + Security + Deps + Types                                                        | `workflows/tribunal-backend.md`     |
 | `/tribunal-frontend`    | Logic + Security + Frontend + Types                                                    | `workflows/tribunal-frontend.md`    |
 | `/tribunal-database`    | Logic + Security + SQL                                                                 | `workflows/tribunal-database.md`    |
@@ -71,7 +71,7 @@ Type any of these in your AI IDE chat:
 
 ---
 
-## The 18 Tribunal Agents
+## The 19 Tribunal Agents
 
 | Agent                    | File                               | Activates When                                                                      |
 | ------------------------ | ---------------------------------- | ----------------------------------------------------------------------------------- |
@@ -267,16 +267,15 @@ Script failures follow cascade rules:
 .agent/
 ├── ARCHITECTURE.md          ← This file
 ├── GEMINI.md                ← Root behavior config (includes /swarm routing)
-├── agents/                  ← 43 specialist + reviewer agents (18 reviewers + 25 domain)
+├── agents/                  ← 43 specialist + reviewer agents (19 reviewers + 24 domain)
 │   ├── supervisor-agent.md  ← Swarm triage, dispatch, synthesis
 │   ├── swarm-worker-contracts.md  ← WorkerRequest/WorkerResult schemas
 │   └── swarm-worker-registry.md   ← Task type → agent routing map
 ├── rules/GEMINI.md          ← Master rules (P0 priority)
-├── scripts/                 ← 25 Python/JS automation scripts
-│   └── swarm_dispatcher.js  ← Validates WorkerRequest/WorkerResult JSON
-├── skills/                  ← 90 modular skill packages (all hardened)
+├── scripts/                 ← 28 Python/JS automation scripts
+├── skills/                  ← 106 modular skill packages (all hardened)
 ├── patterns/                ← 5 ADK skill base patterns
 ├── history/                 ← Case Law + Skill Evolution data (user-generated, preserved on update)
-└── workflows/               ← 31 slash command definitions
+└── workflows/               ← 34 slash command definitions
     └── swarm.md             ← /swarm orchestration procedure
 ```

package/.agent/agents/frontend-reviewer.md

@@ -159,4 +159,17 @@ const { id } = await params;
 
 ---
 
+## Section 7: Fabel Design Standards
+
+### Platform-Aware Rendering Checks
+- Verify if the component differentiates between desktop and mobile targets where applicable.
+- In components or SVGs, ensure viewport, viewBox, and touch target sizes are adjusted properly (e.g., minimum 44x44px touch targets on mobile).
+- Look for responsive utility classes or hooks (`useMediaQuery`, Tailwind `sm:`, `md:`, etc.) to verify adaptation logic.
+
+### Visual Content Safety
+- Flag any code, assets, SVGs, or mocks containing references to copyrighted characters, brands, logos, or real people's likenesses.
+- Ensure only generic symbols, standard library icons (e.g., Lucide React), or explicitly clean SVGs are used.
+
+---
+
 ---

package/.agent/agents/frontend-specialist.md

@@ -255,3 +255,17 @@ Every component I generate meets WCAG 2.2 AA:
 - **Color contrast**: Text minimum 4.5:1 (AA) on its background
 
 ---
+
+## 9. Fabel Output Format & Structure Rules
+
+### File Creation Decision Tree
+- **Inline Output:** Code changes under 20 lines should be provided inline.
+- **File Output:** Code changes over 20 lines, new components, or major refactors must be written directly to a file rather than printed in the response.
+- **Documentation:** Create markdown files in specific doc folders rather than verbose text walls in chat.
+
+### Formatting & Prose Discipline
+- **Prose-First:** Use paragraph prose as the default form of communication.
+- **Anti-Slop:** Avoid bulleted list lists and nested headers for short points. Use bullets only when describing 4 or more distinct, multifaceted items.
+- **No Decorative Sections:** Eliminate filler sections, conversational headers, or repetitive explanations. Keep it scannable, dense, and premium.
+
+---

package/.agent/agents/logic-reviewer.md

@@ -115,4 +115,15 @@ Flag any:
 
 ---
 
+## Fabel Epistemic Standards
+
+### Stale Context Detection
+- Flag any code that modifies a file or calls a method based on an out-of-date or assumed file state. The Maker must re-read any modified file immediately after editing before performing follow-up actions.
+
+### // VERIFY: [reason] Enforcement
+- Flag any lines calling an undocumented API, third-party library method, or complex framework feature that lack an explicit `// VERIFY: [reason]` comment explaining why the Maker believes the call is correct.
+- Ensure the verification reason is specific, not generic (e.g., `// VERIFY: Check if package.json has this version`).
+
+---
+
 ---

package/.agent/agents/orchestrator.md

@@ -185,3 +185,18 @@ Approve?  Y = write to disk | N = discard | R = revise with feedback
 **Nothing is written to disk without explicit human approval.**
 
 ---
+
+## 9. Fabel Tool Selection & Scaling Rules
+
+### Tool Priority Hierarchy
+1. **Workspace Tools** — prioritize local file search, grep, view, and edit. These are fast and conserve token space.
+2. **Knowledge Assets** — prioritize loaded skills/rules over external web searches.
+3. **Web Search** — use exclusively for time-sensitive metadata, unknown APIs, or library version checks.
+
+### Complexity-Scaled budgets
+- **Simple Lookup / Fact:** 1 tool call max.
+- **Single-File Edit / Bug Fix:** 2–4 tool calls.
+- **Complex Implementation:** 5–10 tool calls.
+- **Maximum Threshold:** If a task requires >20 tool calls, the Orchestrator MUST halt and request the user to decompose the task or provide guidance.
+
+---

package/.agent/agents/security-auditor.md

@@ -178,3 +178,16 @@ app.use((err: Error, req, res, next) => {
 ```
 
 ---
+
+## 4. Prompt Injection Defense & LLM Call Security
+
+### Prompt Injection Defenses
+- Ensure user-provided input is never concatenated directly into top-level system prompts.
+- Verify that user inputs are isolated in `<user_provided_context>` or custom XML tags, and system prompts explicitly instruct the model to ignore instructions within those tags.
+- Verify that HTML/XML tags are stripped or sanitized from user input before sending to LLM APIs.
+
+### LLM Call Audit Checklist
+- Flag any hardcoded API keys. They must be loaded from environment variables.
+- Verify model strings against official current lists (no invented names like `claude-4-opus` or `gpt-5`).
+- Ensure all LLM API calls are asynchronous and properly awaited.
+- Check that output parsing is wrapped in try-catch blocks and parses structured JSON safely.

package/.agent/agents/ui-ux-auditor.md

@@ -128,7 +128,7 @@ Required fix: [concrete action the Maker Agent must take]
 
 ```
 ✅ 8pt grid system: 8px, 16px, 24px, 32px, 48px, 64px, 96px, 128px
-✅ CSS custom properties: --space-2: 8px, --space-4: 16px, --space-6: 24px
+✅ Custom CSS properties: --space-2: 8px, --space-4: 16px, --space-6: 24px
 ✅ Section padding: 80px–120px vertical on desktop, 48px–64px on mobile
 ✅ Component padding: consistent horizontal padding on all containers
 ```
@@ -224,36 +224,7 @@ Required fix: [concrete action the Maker Agent must take]
 
 ```
 COLOR
-□ Primary color is not purple/violet
-□ Using CSS custom properties, not raw hex values
-□ Dark mode background is near-black (not generic gray-900)
-□ Contrast ratios meet WCAG AA
-
-TYPOGRAPHY
-□ Fluid sizing with clamp() for responsive text
-□ Clear hierarchy: hero / heading / subheading / body / caption
-□ Max line length bounded for readability
-
-MOTION
-□ Entrance animation present on key elements
-□ Hover states on all interactive elements
-□ @media (prefers-reduced-motion) implemented
-□ No linear easing — use cubic-bezier throughout
-
-LAYOUT
-□ Not a standard hero layout without justification
-□ Not a bento grid without justification
-□ 8pt spacing grid used consistently
-
-TEXTURE
-□ Background has depth: grain, shadow, or layering
-□ No flat mesh gradient backgrounds
-□ Glassmorphism used sparingly, not as the default
-
-ACCESSIBILITY
-□ Focus-visible styles on all interactive elements
-□ Semantic HTML (button not div, nav not ul, etc.)
-□ ARIA labels where needed for icon-only buttons
+... [truncated for spacing] ...
 ```
 
 ---
@@ -297,4 +268,9 @@ VBC PROTOCOL (Verdict-Based Correction):
 □ Every REJECTED verdict includes a concrete "Required fix" with code
 □ Every WARNING includes a code example from the positive patterns above
 □ No vague feedback — every verdict is actionable
+
+FABEL DESIGN INTELLIGENCE:
+□ Did I check that the visual output adapts to the target platform (mobile vs desktop)?
+□ Are touch targets at least 44x44px for touch screens?
+□ Did I verify that there are no copyrighted characters, logos, or real people in the generated UI?
 ```