agent-toolbelt 0.2.2 → 0.2.3

package/README.md

@@ -2,7 +2,7 @@
 
 Official SDK for [Agent Toolbelt](https://agent-toolbelt-production.up.railway.app) — a suite of focused API tools for AI agents and developers.
 
-**Typed client + LangChain tool wrappers** for schema generation, text extraction, token counting, CSV conversion, Markdown conversion, URL metadata, regex building, cron expressions, address normalization, color palette generation, and brand kit creation.
+**Typed client + LangChain tool wrappers** for schema generation, text extraction, token counting, CSV conversion, Markdown conversion, URL metadata, regex building, cron expressions, address normalization, color palette generation, brand kit creation, meeting action item extraction, and prompt optimization.
 
 ## Install
 
@@ -121,6 +121,12 @@ await client.colorPalette({ description: "calm fintech blue", count: 5 });
 
 // Generate a full brand kit — colors, typography, CSS/Tailwind tokens
 await client.brandKit({ name: "Solaris Health", industry: "healthcare", vibe: ["modern", "trustworthy"], format: "full" });
+
+// Extract action items, decisions, and summary from meeting notes
+await client.meetingActionItems({ notes: transcript, format: "full", participants: ["Sarah", "John"] });
+
+// Analyze and improve an LLM prompt
+await client.promptOptimizer({ prompt: "Summarize this and tell me the main points.", model: "gpt-4o", mode: "both" });
 ```
 
 ---
@@ -163,6 +169,9 @@ const result = await agent.invoke({
 | `normalize_address` | Normalize US addresses to USPS format |
 | `generate_color_palette` | Generate color palettes with WCAG scores and CSS variables |
 | `generate_brand_kit` | Generate full brand kit — colors, typography, CSS/Tailwind tokens |
+| `strip_image_metadata` | Strip EXIF/GPS/IPTC/XMP metadata from images for privacy |
+| `extract_meeting_action_items` | Extract action items, decisions, and summary from meeting notes |
+| `optimize_prompt` | Analyze and improve LLM prompts with scores and rewrite |
 
 ---
 
@@ -181,6 +190,9 @@ const result = await agent.invoke({
 | `address-normalizer` | $0.0005 / call |
 | `color-palette` | $0.0005 / call |
 | `brand-kit` | $0.001 / call |
+| `image-metadata-stripper` | $0.001 / call |
+| `meeting-action-items` | $0.05 / call |
+| `prompt-optimizer` | $0.05 / call |
 
 ## License
 

package/dist/index.d.mts

@@ -185,6 +185,26 @@ interface BrandKitResult {
         googleFontsUrl?: string;
     };
 }
+interface PromptOptimizerResult {
+    mode: "improve" | "analyze" | "both";
+    model: string;
+    scores?: {
+        clarity: number;
+        specificity: number;
+        structure: number;
+        completeness: number;
+        overall: number;
+    };
+    issues?: string[];
+    suggestions?: string[];
+    improvedPrompt?: string;
+    changesSummary?: string[];
+    tokenStats: {
+        original: number;
+        improved?: number;
+        delta?: number;
+    };
+}
 interface MeetingActionItem {
     id: number;
     owner: string;
@@ -294,6 +314,13 @@ declare class AgentToolbelt {
         format?: "hex" | "rgb" | "hsl" | "all";
         includeShades?: boolean;
     }): Promise<ColorPaletteResult>;
+    /** Analyze and improve an LLM prompt — scores clarity, specificity, structure, and completeness */
+    promptOptimizer(input: {
+        prompt: string;
+        model?: string;
+        task?: string;
+        mode?: "improve" | "analyze" | "both";
+    }): Promise<PromptOptimizerResult>;
     /** Extract action items, decisions, and summary from meeting notes or transcripts */
     meetingActionItems(input: {
         notes: string;
@@ -316,4 +343,4 @@ declare class AgentToolbelt {
     }): Promise<BrandKitResult>;
 }
 
-export { type AddressNormalizerResult, AgentToolbelt, type AgentToolbeltOptions, type BrandKitResult, type ColorPaletteResult, type CronBuilderResult, type CsvToJsonResult, type ImageMetadataStripperResult, type MarkdownConverterResult, type MeetingActionItem, type MeetingActionItemsResult, type RegexBuilderResult, type SchemaGeneratorResult, type TextExtractorResult, type TokenCountResult, type UrlMetadataResult };
+export { type AddressNormalizerResult, AgentToolbelt, type AgentToolbeltOptions, type BrandKitResult, type ColorPaletteResult, type CronBuilderResult, type CsvToJsonResult, type ImageMetadataStripperResult, type MarkdownConverterResult, type MeetingActionItem, type MeetingActionItemsResult, type PromptOptimizerResult, type RegexBuilderResult, type SchemaGeneratorResult, type TextExtractorResult, type TokenCountResult, type UrlMetadataResult };

package/dist/index.d.ts

@@ -185,6 +185,26 @@ interface BrandKitResult {
         googleFontsUrl?: string;
     };
 }
+interface PromptOptimizerResult {
+    mode: "improve" | "analyze" | "both";
+    model: string;
+    scores?: {
+        clarity: number;
+        specificity: number;
+        structure: number;
+        completeness: number;
+        overall: number;
+    };
+    issues?: string[];
+    suggestions?: string[];
+    improvedPrompt?: string;
+    changesSummary?: string[];
+    tokenStats: {
+        original: number;
+        improved?: number;
+        delta?: number;
+    };
+}
 interface MeetingActionItem {
     id: number;
     owner: string;
@@ -294,6 +314,13 @@ declare class AgentToolbelt {
         format?: "hex" | "rgb" | "hsl" | "all";
         includeShades?: boolean;
     }): Promise<ColorPaletteResult>;
+    /** Analyze and improve an LLM prompt — scores clarity, specificity, structure, and completeness */
+    promptOptimizer(input: {
+        prompt: string;
+        model?: string;
+        task?: string;
+        mode?: "improve" | "analyze" | "both";
+    }): Promise<PromptOptimizerResult>;
     /** Extract action items, decisions, and summary from meeting notes or transcripts */
     meetingActionItems(input: {
         notes: string;
@@ -316,4 +343,4 @@ declare class AgentToolbelt {
     }): Promise<BrandKitResult>;
 }
 
-export { type AddressNormalizerResult, AgentToolbelt, type AgentToolbeltOptions, type BrandKitResult, type ColorPaletteResult, type CronBuilderResult, type CsvToJsonResult, type ImageMetadataStripperResult, type MarkdownConverterResult, type MeetingActionItem, type MeetingActionItemsResult, type RegexBuilderResult, type SchemaGeneratorResult, type TextExtractorResult, type TokenCountResult, type UrlMetadataResult };
+export { type AddressNormalizerResult, AgentToolbelt, type AgentToolbeltOptions, type BrandKitResult, type ColorPaletteResult, type CronBuilderResult, type CsvToJsonResult, type ImageMetadataStripperResult, type MarkdownConverterResult, type MeetingActionItem, type MeetingActionItemsResult, type PromptOptimizerResult, type RegexBuilderResult, type SchemaGeneratorResult, type TextExtractorResult, type TokenCountResult, type UrlMetadataResult };

package/dist/index.js

@@ -90,6 +90,10 @@ var AgentToolbelt = class {
   colorPalette(input) {
     return this.call("color-palette", input);
   }
+  /** Analyze and improve an LLM prompt — scores clarity, specificity, structure, and completeness */
+  promptOptimizer(input) {
+    return this.call("prompt-optimizer", input);
+  }
   /** Extract action items, decisions, and summary from meeting notes or transcripts */
   meetingActionItems(input) {
     return this.call("meeting-action-items", input);

package/dist/index.mjs

@@ -64,6 +64,10 @@ var AgentToolbelt = class {
   colorPalette(input) {
     return this.call("color-palette", input);
   }
+  /** Analyze and improve an LLM prompt — scores clarity, specificity, structure, and completeness */
+  promptOptimizer(input) {
+    return this.call("prompt-optimizer", input);
+  }
   /** Extract action items, decisions, and summary from meeting notes or transcripts */
   meetingActionItems(input) {
     return this.call("meeting-action-items", input);

package/dist/langchain.js

@@ -165,6 +165,21 @@ function createLangChainTools(client) {
         return JSON.stringify(result);
       }
     }),
+    // ---- Prompt Optimizer ----
+    new import_tools.DynamicStructuredTool({
+      name: "optimize_prompt",
+      description: "Analyze and improve an LLM prompt for clarity, specificity, structure, and completeness. Use this when a prompt isn't performing well, to prepare prompts before deploying them, or to learn what makes a good prompt for a specific model. Returns scores (1-10) for each quality dimension, a list of issues found, an improved rewrite, and a summary of what changed and why. Supports targeting specific models: gpt-4o, claude-3-5-sonnet, gpt-3.5-turbo, etc.",
+      schema: import_zod.z.object({
+        prompt: import_zod.z.string().describe("The LLM prompt to analyze and/or improve"),
+        model: import_zod.z.string().default("gpt-4o").describe("Target model to optimize for"),
+        task: import_zod.z.string().optional().describe("What this prompt is trying to accomplish (helps generate targeted suggestions)"),
+        mode: import_zod.z.enum(["improve", "analyze", "both"]).default("both").describe("'both' returns analysis + improved prompt; 'analyze' scores only; 'improve' rewrites only")
+      }),
+      func: async ({ prompt, model, task, mode }) => {
+        const result = await client.promptOptimizer({ prompt, model, task, mode });
+        return JSON.stringify(result);
+      }
+    }),
     // ---- Meeting Action Items ----
     new import_tools.DynamicStructuredTool({
       name: "extract_meeting_action_items",

package/dist/langchain.mjs

@@ -141,6 +141,21 @@ function createLangChainTools(client) {
         return JSON.stringify(result);
       }
     }),
+    // ---- Prompt Optimizer ----
+    new DynamicStructuredTool({
+      name: "optimize_prompt",
+      description: "Analyze and improve an LLM prompt for clarity, specificity, structure, and completeness. Use this when a prompt isn't performing well, to prepare prompts before deploying them, or to learn what makes a good prompt for a specific model. Returns scores (1-10) for each quality dimension, a list of issues found, an improved rewrite, and a summary of what changed and why. Supports targeting specific models: gpt-4o, claude-3-5-sonnet, gpt-3.5-turbo, etc.",
+      schema: z.object({
+        prompt: z.string().describe("The LLM prompt to analyze and/or improve"),
+        model: z.string().default("gpt-4o").describe("Target model to optimize for"),
+        task: z.string().optional().describe("What this prompt is trying to accomplish (helps generate targeted suggestions)"),
+        mode: z.enum(["improve", "analyze", "both"]).default("both").describe("'both' returns analysis + improved prompt; 'analyze' scores only; 'improve' rewrites only")
+      }),
+      func: async ({ prompt, model, task, mode }) => {
+        const result = await client.promptOptimizer({ prompt, model, task, mode });
+        return JSON.stringify(result);
+      }
+    }),
     // ---- Meeting Action Items ----
     new DynamicStructuredTool({
       name: "extract_meeting_action_items",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agent-toolbelt",
-  "version": "0.2.2",
+  "version": "0.2.3",
   "description": "Official SDK for Agent Toolbelt — typed API client and LangChain tool wrappers for schema generation, text extraction, token counting, CSV conversion, and more.",
   "main": "./dist/index.js",
   "module": "./dist/index.mjs",