Security Scanner MCP has 5 tools with mixed quality. Tool names follow verb_noun convention (scan_url, list_recent_scans, get_scan, scan_body, clear_scans), which is good. However, descriptions are in Turkish, creating a language barrier for English-speaking LLM agents. Descriptions exist but lack clarity on when to use each tool vs. alternatives. Input schemas are present with types and enums, but parameter descriptions are sparse or missing context. scan_url has 4 params with descriptions, but credentialCheck and async lack detail on side effects. list_recent_scans and get_scan have minimal schemas. clear_scans is destructive but lacks confirmation/dry-run pattern. No output schemas documented. Error handling is absent, no guidance on retryability, user-fixable errors, or recovery paths.

• Translate all tool and parameter descriptions to English. Replace Turkish text with clear, LLM-optimized English descriptions (50 - 200 chars per tool, 30 - 80 chars per param).
• Document output schemas for all tools. Specify what fields each tool returns (e.g., list_recent_scans returns array of {url, mode, score, completedAt, scanId}). This enables tool chaining.
• Add confirmation or dry-run support to clear_scans. Require an explicit 'confirm=true' parameter or offer a 'dry_run' mode that lists what would be deleted without deleting.
• Expand parameter descriptions for scan_url: clarify async polling (how to check status, timeout), explain credentialCheck side effects, and document ALLOWED_ACTIVE_HOSTS allowlist behavior.
• Add error handling guidance: document which errors are retryable (network timeout, rate limit) vs. user-fixable (invalid URL, host not allowed) vs. fatal (internal server error). Include recovery hints.