https://cve.radiocsirt.org Contains only the most 10 recent entries. http://www.rssboard.org/rss-specification python-feedgen en Sun, 04 Oct 2026 21:06:51 +0000 https://cve.radiocsirt.org/vuln/cve-2026-44661 <p><strong>Affected:</strong> universal-tool-calling-protocol python-utcp</p> <p>python-utcp is the python implementation of UTCP. Prior to 1.1.3, the utcp-http plugin is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. register_manual() validates the discovery URL against an HTTPS / loopback allowlist, but call_tool() and call_tool_streaming() reuse the resolved tool_call_template.url directly without revalidating, and the OpenAPI converter blindly trusts whatever servers[0].url an attacker-hosted spec declares. An attacker who hosts a malicious OpenAPI spec on a legitimate HTTPS endpoint can declare e.g. servers: [{ url: &#34;http://127.0.0.1:9090&#34; }] or servers: [{ url: &#34;http://169.254.169.254&#34; }]; the OpenAPI converter then produces tools whose URL points at internal services on the agent host. All three HTTP-class protocols (utcp_http.http, utcp_http.streamable_http, utcp_http.sse) shared the same gap. This vulnerability is fixed in 1.1.3.</p> <p><strong>Affected:</strong> universal-tool-calling-protocol python-utcp</p> <p>python-utcp is the python implementation of UTCP. Prior to 1.1.3, the utcp-http plugin is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. register_manual() validates the discovery URL against an HTTPS / loopback allowlist, but call_tool() and call_tool_streaming() reuse the resolved tool_call_template.url directly without revalidating, and the OpenAPI converter blindly trusts whatever servers[0].url an attacker-hosted spec declares. An attacker who hosts a malicious OpenAPI spec on a legitimate HTTPS endpoint can declare e.g. servers: [{ url: &#34;http://127.0.0.1:9090&#34; }] or servers: [{ url: &#34;http://169.254.169.254&#34; }]; the OpenAPI converter then produces tools whose URL points at internal services on the agent host. All three HTTP-class protocols (utcp_http.http, utcp_http.streamable_http, utcp_http.sse) shared the same gap. This vulnerability is fixed in 1.1.3.</p> https://cve.radiocsirt.org/vuln/cve-2026-44661 https://cve.radiocsirt.org/vuln/ghsa-39j6-4867-gg4w <p><strong>Affected:</strong> PyPI: utcp-http</p> <p>## Summary</p> <p>The `utcp-http` plugin is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. `register_manual()` validates the discovery URL against an HTTPS / loopback allowlist, but `call_tool()` and `call_tool_streaming()` reuse the resolved `tool_call_template.url` directly without revalidating. An attacker who hosts a malicious OpenAPI spec on a legitimate HTTPS endpoint can declare `servers: [{ url: &#34;http://169.254.169.254&#34; }]` (or any internal address) in the spec; the OpenAPI converter blindly trusts that value and the tool becomes a blind SSRF primitive that exposes cloud metadata, internal services, and other firewalled-only endpoints to the LLM caller.</p> <p>All three HTTP-class protocols (`utcp_http.http`, `utcp_http.streamable_http`, `utcp_http.sse`) shared the same gap, plus a separate prefix-bypass: the previous `startswith(&#34;http://localhost&#34;)` check let URLs like `http://localhost.evil.com` through.</p> <p>## Impact</p> <p>A remote attacker who can convince the agent (via the LLM context, prompt injection, or a tool-discovery surface) to register their HTTPS OpenAPI URL can: - Map internal networks behind the agent. - Read AWS/GCP IAM credentials from cloud metadata endpoints (`http://169.254.169.254`, `http://metadata.google.internal`). - Reach unauthenticated internal services (Elasticsearch, Redis HTTP, internal admin panels). - Have responses returned to the LLM, which combined with p…</p> <p><strong>Affected:</strong> PyPI: utcp-http</p> <p>## Summary</p> <p>The `utcp-http` plugin is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. `register_manual()` validates the discovery URL against an HTTPS / loopback allowlist, but `call_tool()` and `call_tool_streaming()` reuse the resolved `tool_call_template.url` directly without revalidating. An attacker who hosts a malicious OpenAPI spec on a legitimate HTTPS endpoint can declare `servers: [{ url: &#34;http://169.254.169.254&#34; }]` (or any internal address) in the spec; the OpenAPI converter blindly trusts that value and the tool becomes a blind SSRF primitive that exposes cloud metadata, internal services, and other firewalled-only endpoints to the LLM caller.</p> <p>All three HTTP-class protocols (`utcp_http.http`, `utcp_http.streamable_http`, `utcp_http.sse`) shared the same gap, plus a separate prefix-bypass: the previous `startswith(&#34;http://localhost&#34;)` check let URLs like `http://localhost.evil.com` through.</p> <p>## Impact</p> <p>A remote attacker who can convince the agent (via the LLM context, prompt injection, or a tool-discovery surface) to register their HTTPS OpenAPI URL can: - Map internal networks behind the agent. - Read AWS/GCP IAM credentials from cloud metadata endpoints (`http://169.254.169.254`, `http://metadata.google.internal`). - Reach unauthenticated internal services (Elasticsearch, Redis HTTP, internal admin panels). - Have responses returned to the LLM, which combined with p…</p> https://cve.radiocsirt.org/vuln/ghsa-39j6-4867-gg4w