feat: HTTP Tool Security Hardening — SSRF Protection, Response Size Limits & Input Validation

[esafwan]

Summary

This is a clean implementation of the HTTP tool security review and hardening from #222. All changes have been applied to a fresh branch off the current develop.

Changes Implemented

1. SSRF Protection & Network Hardening (huf/ai/http_handler.py)

• DNS-resolving IP check: Blocks all private/internal network addresses (RFC 1918 ranges, localhost 127.0.0.1, 0.0.0.0, ::1, and cloud metadata IPs 169.254.169.254). Relies on socket-resolved IPs to prevent DNS rebinding attacks.
• Method Allowlist: Restricts outbound HTTP requests to safe/standard methods: GET, POST, PUT, PATCH, DELETE, HEAD.
• Response Payload Cap: Enforces a strict 10MB limit on the total response body size. Pre-checks the Content-Length header if available, and measures chunks iteratively during streaming to stop downloads immediately when the limit is breached.

2. Broad Ecosystem SSRF Guarding

• URL Attachment SSRF Guard (huf/ai/tool_functions.py): Integrates validate_url() check inside _download_content() to stop agents downloading files from private endpoints.
• URL Knowledge Extractor (huf/ai/knowledge/extractors/url.py): Validates user-supplied knowledge source URLs.
• Image Generation Attachment Guard (huf/ai/sdk_tools.py): Validates generated or fetched image URLs.

3. Server-Side Validation (huf/huf/doctype/agent_tool_function/agent_tool_function.py)

• CRLF Header Injection Prevention: Ensures that all user-supplied headers do not contain carriage returns (\r) or line feeds (\n).
• Reserved Parameter Collision Guard: Rejects parameters named url, params, json_data, headers, data, tool_name, or method to prevent conflicts with HTTP handler kwargs.
• Base URL Verification: Enforces that the base_url is a valid http or https URL with a proper hostname.

4. Frontend Type Safety & Validation (frontend/src/components/tools/toolCreationForm.utils.ts)

• Zod Schema Updates: Validates the base URL as a valid http/https format, and properly types HTTP headers array.

Verification & Testing

• SSRF block verified: requests to http://127.0.0.1:8000 or local URLs are successfully blocked.
• Public outbound requests work: https://httpbin.org/get successfully completed with a 200 response.
• Frontend builds cleanly without TypeScript compiler issues.

Closes #222