diff --git a/prompts/dev/add-keyword.md b/prompts/dev/add-keyword.md
new file mode 100644
index 00000000..5aa90571
--- /dev/null
+++ b/prompts/dev/add-keyword.md
@@ -0,0 +1,151 @@
+Create a new Rhai custom keyword implementation with these specifications:
+
+- When generating new Rhai keyword handlers in Rust, always design them to be thread-safe and fully compatible with async execution. Use Arc for shared state, perform heavy or async operations with tokio::task::block_in_place and Handle::current().block_on, and return results as rhai::Dynamic to maintain safe cross-thread communication between Rust and the Rhai engine.
+
+1. DATABASE REQUIREMENTS:
+- No enums in database schema (only in Rust code)
+- Use direct integer values for enum variants in queries
+- Follow existing connection pooling pattern with AppState
+- Include proper error handling and logging
+
+2. RUST IMPLEMENTATION:
+- Enum definition (Rust-only, no DB enum):
+```rust
+#[repr(i32)]
+pub enum KeywordAction {
+    Action1 = 0,
+    Action2 = 1,
+    Action3 = 2
+}
+```
+
+3. KEYWORD TEMPLATE:
+```rust
+pub fn {keyword_name}_keyword(state: &AppState, engine: &mut Engine) {
+    let db = state.db_custom.clone();
+
+    engine.register_custom_syntax(
+        {syntax_pattern},
+        {is_raw},
+        {
+            let db = db.clone();
+            move |context, inputs| {
+                // Input processing
+                {input_processing}
+
+                let binding = db.as_ref().unwrap();
+                let fut = execute_{keyword_name}(binding, {params});
+
+                let result = tokio::task::block_in_place(||
+                    tokio::runtime::Handle::current().block_on(fut))
+                        .map_err(|e| format!("DB error: {}", e))?;
+
+                {result_handling}
+            }
+        }
+    ).unwrap();
+}
+
+pub async fn execute_{keyword_name}(
+    pool: &PgPool,
+    {params_with_types}
+) -> Result<Value, Box<dyn std::error::Error>> {
+    info!("Executing {keyword_name} with: {debug_params}");
+
+    let result = sqlx::query(
+        "{sql_query_with_i32_enum}"
+    )
+    .bind({enum_value} as i32)
+    {additional_binds}
+    .execute(pool)
+    .await?;
+
+    Ok(json!({
+        "command": "{keyword_name}",
+        {result_fields}
+        "rows_affected": result.rows_affected()
+    }))
+}
+```
+
+4. EXAMPLE IMPLEMENTATION (SET SCHEDULE):
+```rust
+// Enum (Rust-only)
+#[repr(i32)]
+pub enum TriggerKind {
+    Scheduled = 0,
+    TableUpdate = 1,
+    TableInsert = 2,
+    TableDelete = 3
+}
+
+// Keyword implementation
+pub fn set_schedule_keyword(state: &AppState, engine: &mut Engine) {
+    let db = state.db_custom.clone();
+
+    engine.register_custom_syntax(
+        ["SET", "SCHEDULE", "$string$"],
+        true,
+        {
+            let db = db.clone();
+            move |context, inputs| {
+                let cron = context.eval_expression_tree(&inputs[0])?.to_string();
+                let script_name = format!("cron_{}.rhai", cron.replace(' ', "_"));
+
+                let binding = db.as_ref().unwrap();
+                let fut = execute_set_schedule(binding, &cron, &script_name);
+
+                let result = tokio::task::block_in_place(||
+                    tokio::runtime::Handle::current().block_on(fut))
+                        .map_err(|e| format!("DB error: {}", e))?;
+
+                if let Some(rows_affected) = result.get("rows_affected") {
+                    Ok(Dynamic::from(rows_affected.as_i64().unwrap_or(0)))
+                } else {
+                    Err("No rows affected".into())
+                }
+            }
+        }
+    ).unwrap();
+}
+
+pub async fn execute_set_schedule(
+    pool: &PgPool,
+    cron: &str,
+    script_name: &str,
+) -> Result<Value, Box<dyn std::error::Error>> {
+    info!("Executing schedule: {}, {}", cron, script_name);
+
+    let result = sqlx::query(
+        "INSERT INTO system_automations
+        (kind, schedule, script_name)
+        VALUES ($1, $2, $3)"
+    )
+    .bind(TriggerKind::Scheduled as i32)
+    .bind(cron)
+    .bind(script_name)
+    .execute(pool)
+    .await?;
+
+    Ok(json!({
+        "command": "set_schedule",
+        "schedule": cron,
+        "script_name": script_name,
+        "rows_affected": result.rows_affected()
+    }))
+}
+```
+
+5. ADDITIONAL REQUIREMENTS:
+- Maintain consistent tokio runtime handling
+- Include parameter validation
+- Follow existing JSON response format
+- Ensure proper script name generation
+- Include debug logging for all operations
+
+6. OUTPUT FORMAT:
+Provide complete implementation with:
+1. Rust enum definition
+2. Keyword registration function
+3. Execution function
+4. Example usage in Rhai
diff --git a/prompts/dev/add-model.md b/prompts/dev/add-model.md
new file mode 100644
index 00000000..c3c462ef
--- /dev/null
+++ b/prompts/dev/add-model.md
@@ -0,0 +1,94 @@
+
+Create a Rust data model for database storage with optimal size and performance characteristics. Follow these specifications:
+
+**REQUIREMENTS:**
+1. Use appropriate integer types (i32, i16, i8, etc.) based on expected value ranges
+2. Use `Option<T>` for nullable fields to avoid memory overhead
+3. Use `Vec<u8>` for binary data instead of strings when appropriate
+4. Prefer enum representations as integers rather than strings
+5. Use `chrono::DateTime<Utc>` for timestamps
+6. Use `uuid::Uuid` for unique identifiers
+7. Implement necessary traits: `Debug`, `Clone`, `Serialize`, `Deserialize`, `FromRow`
+8. Include validation where appropriate
+9. Consider database index strategy in field design
+
+**CONTEXT:**
+- Database: PostgreSQL/SQLx compatible
+- Serialization: Serde for JSON
+- ORM: SQLx for database operations
+
+**OUTPUT FORMAT:**
+Provide the complete Rust struct with:
+- Struct definition with fields
+- Enum definitions with integer representations
+- Conversion implementations
+- Basic validation if needed
+
+**EXAMPLE REFERENCE:**
+```rust
+use chrono::{DateTime, Utc};
+use serde::{Deserialize, Serialize};
+use sqlx::FromRow;
+use uuid::Uuid;
+
+#[derive(Debug, Clone, Copy, PartialEq)]
+pub enum Status {
+    Pending = 0,
+    Active = 1,
+    Inactive = 2,
+}
+
+impl Status {
+    pub fn from_i16(value: i16) -> Option<Self> {
+        match value {
+            0 => Some(Self::Pending),
+            1 => Some(Self::Active),
+            2 => Some(Self::Inactive),
+            _ => None,
+        }
+    }
+}
+
+#[derive(Debug, FromRow, Serialize, Deserialize)]
+pub struct User {
+    pub id: Uuid,
+    pub status: i16, // Using i16 for enum storage
+    pub email: String,
+    pub age: Option<i16>, // Nullable small integer
+    pub metadata: Vec<u8>, // Binary data for flexibility
+    pub created_at: DateTime<Utc>,
+}
+```
+
+Generate a similar model for: [YOUR DOMAIN HERE]
+```
+
+## Specialized Variants
+
+### For High-Performance Applications
+```
+Add these additional requirements:
+- Use `#[repr(u8)]` for enums to ensure minimal size
+- Consider `Box<str>` instead of `String` for reduced heap overhead
+- Use `arrayvec::ArrayString` for fixed-size short strings
+- Implement `PartialEq` and `Eq` for hash-based operations
+- Include `#[derive(Default)]` where appropriate
+```
+
+### For Embedded/Memory-Constrained Systems
+```
+Add these constraints:
+- Prefer `i16` over `i32` where possible
+- Use `u32` instead of `Uuid` if sequential IDs are acceptable
+- Consider `bitflags` for multiple boolean flags in single byte
+- Use `smol_str::SmolStr` for string optimization
+- Avoid `Vec` in favor of arrays with capacity limits
+```
+
+### For Time-Series Data
+```
+Add time-series specific optimizations:
+- Use `i64` for timestamps as nanoseconds since epoch
+- Use `f32` instead of `f64` for measurements where precision allows
+- Consider `ordered_float::OrderedFloat` for floating-point comparisons
+- Use `#[serde(with = "chrono::serde::ts_seconds")]` for compact serialization
diff --git a/prompts/dev/add-service.md b/prompts/dev/add-service.md
new file mode 100644
index 00000000..002e2286
--- /dev/null
+++ b/prompts/dev/add-service.md
@@ -0,0 +1,57 @@
+Generate a Rust service module following these patterns:
+
+Core Structure:
+
+Use actix-web for HTTP endpoints (get, post, etc.)
+
+Isolate shared resources (DB, clients, config) in AppState
+
+Split logic into reusable helper functions
+
+do not create main logic
+
+Endpoints:
+
+Follow REST conventions (e.g., POST /{resource}/create) use anotations in methods.
+
+Use web::Path for route parameters, web::Json for payloads
+
+Return consistent responses (e.g., HttpResponse::Ok().json(data))
+
+Error Handling:
+
+Wrap fallible operations in Result
+
+Use map_err to convert errors to actix_web::Error
+
+Provide clear error messages (e.g., ErrorInternalServerError)
+
+Async Patterns:
+
+Use async/await for I/O (DB, external APIs)
+
+Leverage streams for pagination/large datasets
+
+Isolate blocking ops in spawn_blocking if needed
+
+Configuration:
+
+Load settings (e.g., URLs, credentials) from AppConfig
+
+Initialize clients (DB, SDKs) at startup (e.g., init_*() helpers)
+
+Documentation:
+
+Add brief doc comments for public functions
+
+Note safety assumptions (e.g., #[post] invariants)
+postgres sqlx
+Omit domain-specific logic (e.g., file/email details), focusing on the scaffolding."
+
+Key Features:
+
+Generic (applies to any service: auth, payments, etc.)
+
+KISS (avoids over-engineering)
+
+Copy-paste friendly (clear patterns without verbosity)
diff --git a/prompts/dev/fix-errors.md b/prompts/dev/fix-errors.md
new file mode 100644
index 00000000..d171d3f5
--- /dev/null
+++ b/prompts/dev/fix-errors.md
@@ -0,0 +1,35 @@
+You are fixing Rust code in a Cargo project. The user is providing problematic code that needs to be corrected.
+
+## Your Task
+Fix ALL compiler errors and logical issues while maintaining the original intent.
+Use Cargo.toml as reference, do not change it.
+Only return input files, all other files already exists.
+If something, need to be added to a external file, inform it separated.
+
+## Critical Requirements
+3. **Respect Cargo.toml** - Check dependencies, editions, and features to avoid compiler errors
+4. **Type safety** - Ensure all types match and trait bounds are satisfied
+5. **Ownership rules** - Fix borrowing, ownership, and lifetime issues
+
+
+MORE RULES:
+- Return only the modified files as a single `.sh` script using `cat`, so the - code can be restored directly.
+- You MUST return exactly this example format:
+```sh
+#!/bin/bash
+
+# Restore fixed Rust project
+
+cat > src/<filenamehere>.rs << 'EOF'
+use std::io;
+
+// test
+
+cat > src/<anotherfile>.rs << 'EOF'
+// Fixed library code
+pub fn add(a: i32, b: i32) -> i32 {
+    a + b
+}
+EOF
+
+----