@johpaz/hive-agents 0.0.37 → 0.0.39

package/packages/core/src/tools/web/browser-type.ts

@@ -53,7 +53,7 @@ export const browserTypeTool: Tool = {
       log.warn("Browser not available");
       return {
         ok: false,
-        error: "Browser automation not available. Install Chrome/Chromium.",
+        error: "Browser automation not available. Install agent-browser.",
       };
     }
 
@@ -61,24 +61,19 @@ export const browserTypeTool: Tool = {
 
     try {
       const view = await browserService.getView();
-      if (!view) return { ok: false, error: "Browser automation not available. Install Chrome/Chromium." };
+      if (!view) return { ok: false, error: "Browser automation not available. Install agent-browser." };
 
       if (url) {
         await view.navigate(url);
         await Bun.sleep(500);
       }
 
-      // click(selector) waits for actionability then focuses the element
-      await view.click(selector, { timeout });
-
       if (clear) {
-        // Ctrl+A → Backspace to clear existing content
-        await view.press("a", { modifiers: ["Control"] });
-        await view.press("Backspace");
+        await (view as any).fill(selector, text);
+      } else {
+        await (view as any).typeIn(selector, text);
       }
 
-      await view.type(text);
-
       const currentUrl = view.url;
       log.info(`Type successful: "${text.substring(0, 50)}${text.length > 50 ? "..." : ""}" into ${selector}`);
 

package/packages/core/src/tools/web/browser-wait.ts

@@ -54,7 +54,7 @@ export const browserWaitTool: Tool = {
       log.warn("Browser not available");
       return {
         ok: false,
-        error: "Browser automation not available. Install Chrome/Chromium.",
+        error: "Browser automation not available. Install agent-browser.",
       };
     }
 
@@ -69,7 +69,7 @@ export const browserWaitTool: Tool = {
 
     try {
       const view = await browserService.getView();
-      if (!view) return { ok: false, error: "Browser automation not available. Install Chrome/Chromium." };
+      if (!view) return { ok: false, error: "Browser automation not available. Install agent-browser." };
 
       if (url) {
         await view.navigate(url);

package/packages/core/src/tools/web/index.ts

@@ -1,7 +1,7 @@
 /**
  * Web Tools - Browser automation + Web utilities
  * 
- * Browser tools use Puppeteer/Chromium (auto-managed).
+ * Browser tools use agent-browser (Rust CLI, auto-managed).
  */
 
 import type { Tool } from "../types.ts";

package/packages/core/src/utils/logger.ts

@@ -1,5 +1,6 @@
 import { mkdirSync, unlinkSync, renameSync, existsSync } from "node:fs";
 import * as path from "node:path";
+import { homedir } from "node:os";
 import { getHiveDir, loadConfig } from "../config/loader.ts";
 
 export type LogLevel = "debug" | "info" | "warn" | "error";
@@ -81,7 +82,7 @@ const COLORS = {
 
 function expandPath(p: string): string {
   if (p.startsWith("~")) {
-    return path.join(process.env.HOME || "", p.slice(1));
+    return path.join(homedir(), p.slice(1));
   }
   return p;
 }

package/packages/mcp/src/manager.ts

@@ -3,6 +3,7 @@ import type { Transport } from "@modelcontextprotocol/sdk/shared/transport.js";
 import type { MCPConfig, MCPServerConfig } from "./config";
 import { logger, type LogHandler } from "./logger";
 import * as path from "node:path";
+import { homedir } from "node:os";
 import {
   createTransport,
   type TransportType,
@@ -139,7 +140,7 @@ export class MCPClientManager {
 
   private expandPath(p: string): string {
     if (p.startsWith("~")) {
-      return path.join(process.env.HOME ?? "", p.slice(1));
+      return path.join(homedir(), p.slice(1));
     }
     return p;
   }

package/packages/skills/src/bundled/api/api_client/SKILL.md

@@ -0,0 +1,132 @@
+---
+name: api_client
+description: "Make HTTP requests to REST APIs using curl-like methods (GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS)"
+version: 1.0.0
+author: Hive Team
+icon: "🌐"
+category: api
+permissions:
+  - internet_access
+dependencies: []
+tools: [api_request]
+
+# Structured skill fields
+triggers:
+  - "llama a la api"
+  - "llama al api"
+  - "consume la api"
+  - "haz una petición"
+  - "haz un request"
+  - "envía un post"
+  - "envía un put"
+  - "envía un delete"
+  - "curl"
+  - "api request"
+  - "rest api"
+  - "endpoint"
+  - "webhook"
+  - "integrar con api"
+  - "conectar con api"
+  - "obtener datos de api"
+  - "enviar datos a api"
+
+preferred_agents: []
+
+steps:
+  - step: 1
+    action: api_request
+    instruction: "Call the API endpoint with the correct HTTP method, headers, and body. Always include Content-Type and Authorization if required by the API."
+    params:
+      method: "GET | POST | PUT | DELETE | PATCH"
+      url: "https://api.example.com/endpoint"
+      headers:
+        Content-Type: "application/json"
+        Authorization: "Bearer TOKEN_IF_NEEDED"
+      body: '{"key":"value"}'
+    output: api_response
+
+  - step: 2
+    action: validate
+    instruction: "Check the response status. If 2xx, extract and present the data. If 4xx/5xx, analyze the error and suggest fixes (missing auth, wrong payload, rate limit)."
+    output: validated_result
+
+rules:
+  - "Always use api_request instead of web_fetch when you need custom headers, body, or non-GET methods"
+  - "Never expose API keys or tokens in the final response to the user — mention they are handled securely"
+  - "If the API returns JSON, parse and present it in a readable format (tables, lists, markdown)"
+  - "For paginated APIs, indicate if more pages exist and offer to fetch them"
+  - "Respect rate limits — if a 429 is returned, wait and retry or inform the user"
+  - "Always include the Content-Type header when sending a body (usually application/json)"
+  - "Use query_params object instead of manually appending to the URL"
+  - "If the user provides a curl command, translate it into api_request parameters"
+
+output_format:
+  structure: markdown
+  sections:
+    - "summary"
+    - "request_details"
+    - "response_data"
+    - "next_steps"
+  max_length: "800 words unless user requests full raw response"
+
+examples:
+  - user_input: "haz un get a https://api.github.com/users/octocat"
+    expected_behavior: "api_request({ method: 'GET', url: 'https://api.github.com/users/octocat' }) → parse JSON → show key fields"
+
+  - user_input: "envía un post a https://httpbin.org/post con {nombre: 'Juan'}"
+    expected_behavior: "api_request({ method: 'POST', url: 'https://httpbin.org/post', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({nombre:'Juan'}) }) → show response"
+
+  - user_input: "actualiza el recurso 123 en la API con {status: 'active'}"
+    expected_behavior: "api_request({ method: 'PUT', url: '.../123', headers: { 'Content-Type': 'application/json', 'Authorization': 'Bearer ...' }, body: JSON.stringify({status:'active'}) }) → confirm update"
+
+  - user_input: "curl -X DELETE https://api.example.com/items/456 -H 'Authorization: Bearer token123'"
+    expected_behavior: "Translate curl to api_request({ method: 'DELETE', url: 'https://api.example.com/items/456', headers: { 'Authorization': 'Bearer token123' } }) → confirm deletion"
+---
+
+# API Client Skill
+
+## Cuándo se Activa
+
+Esta skill se activa cuando el usuario necesita interactuar con una API REST: consultar datos, crear recursos, actualizar, eliminar, o cualquier operación HTTP.
+
+## Herramientas Disponibles
+
+| Tool | Qué hace | Cuándo usarla |
+|------|----------|---------------|
+| `api_request` | Realiza peticiones HTTP completas (GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS) con headers, body, query params y timeout | Siempre que necesites llamar un endpoint REST, webhook, o servicio externo |
+
+## Parámetros de api_request
+
+- `method` (requerido): GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS
+- `url` (requerido): URL completa del endpoint
+- `headers` (opcional): objeto con headers HTTP. Ej: `{ "Authorization": "Bearer TOKEN", "Content-Type": "application/json" }`
+- `body` (opcional): cuerpo de la petición como string. Para JSON, enviar JSON.stringify(objeto)
+- `query_params` (opcional): parámetros de query que se codificarán automáticamente en la URL
+- `timeout_ms` (opcional): timeout en ms. Default: 30000. Máx: 120000
+
+## Diferencia con web_fetch
+
+- `web_fetch`: solo GET, sin headers custom, ideal para scraping de páginas web
+- `api_request`: cualquier método HTTP, headers custom, body, query params — ideal para APIs REST
+
+## Workflow
+
+1. **Identificar endpoint y método** → Determinar URL, método, headers necesarios
+2. **Construir request** → `api_request({ method, url, headers, body })`
+3. **Validar respuesta** → Si 2xx: extraer datos. Si error: analizar y sugerir fix
+4. **Presentar resultados** → JSON parseado en formato legible, no crudo a menos que se pida
+
+## Mejores Prácticas
+
+- Siempre enviar `Content-Type: application/json` cuando el body es JSON
+- Usar `query_params` en lugar de append manual a la URL
+- No exponer tokens/secrets en la respuesta final al usuario
+- Si la API requiere auth, pedirla al usuario o usar variables de entorno
+- Para errores 4xx, revisar: auth, formato del body, campos requeridos, rate limits
+
+## Errores a Evitar
+
+- ❌ Usar web_fetch para POST/PUT/DELETE con headers
+- ❌ Enviar objetos directamente en body (debe ser string)
+- ❌ Olvidar Content-Type al enviar JSON
+- ❌ Exponer API keys en la respuesta visible