Server has 6 tools with explicit schemas and descriptions, but quality is inconsistent. All tools have basic descriptions (10-80 chars), but most are too terse to guide LLM selection effectively. Schemas are present but lack parameter descriptions for nested objects (e.g., 'options' in process_images has no description). No output schemas documented. Error handling is minimal, catch blocks return generic JSON without recovery guidance. Tool naming is reasonable (verb_noun pattern) but lacks clarity on distinctions (scrape_images vs scrape_products vs scrape_product_detail_images are semantically similar). No parameter validation hints, no enum constraints, no pagination support. Security concern: download_images and process_images accept arbitrary file paths without sanitization, risking path traversal attacks.

• Add descriptions to nested 'options' objects: 'removeBg (boolean): Enable background removal using AI model. sharpen (boolean): Apply sharpening filter. sharpenStrength (number, 0-2): Intensity of sharpening (Colab only).'
• Document output schemas for each tool. E.g., scrape_images returns {urls: string[], count: number, timestamp: string}. download_images returns {downloaded: string[], failed: {url: string, reason: string}[], total: number}.
• Rename and clarify scraping tools: 'scrape_page_images' (all images on page), 'scrape_product_list' (product cards with title/price/image), 'scrape_product_detail' (high-res images from single product page). Add 1-2 sentence descriptions explaining use case for each.
• Add path validation: reject paths containing '..', enforce outputDir is writable and within a safe base directory. Return error: 'Invalid path: must be within /safe/base/dir. Got: /etc/passwd'.
• Enhance error responses with recovery hints. E.g., 'Failed to scrape URL: timeout after 30s. Try a different URL or check if the site blocks automated access.' Categorize as retryable, user-fixable, or fatal.