{"id":"CVE-2026-45366","aliases":["EUVD-2026-33053","GHSA-r8j5-8747-88cm"],"status":"published","title":"typescript-utcp: SSRF via attacker-controlled OpenAPI servers[0].url in HTTP communication protocol","description":"typescript-utcp is a typescript implementation of UTCP. Prior to 1.1.2, the @utcp/http package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. registerManual() validates the discovery URL against an HTTPS / loopback allowlist, but callTool() reuses the resolved toolCallTemplate.url directly without revalidating, and the OpenApiConverter 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: \"http://127.0.0.1:9090\" }] or servers: [{ url: \"http://169.254.169.254\" }]; the converter then produces tools whose URL points at internal services on the agent host. This vulnerability is fixed in 1.1.2.","severity":{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-08","score":0.00122,"percentile":0.02005,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-918"],"published":"2026-05-14T20:55:05Z","modified":"2026-09-10T03:50:47Z","fixes":[{"ecosystem":"npm","package":"@utcp/http","fixed_in":["1.1.2"],"sources":["ghsa","osv"]}],"affected":[{"source":"ghsa","ecosystem":"npm","package":"@utcp/http","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"last_affected":"1.1.1"},{"fixed":"1.1.2"}]}]},{"source":"osv","ecosystem":"npm","package":"@utcp/http","ranges":[{"type":"SEMVER","events":[{"introduced":"0"},{"fixed":"1.1.2"}]}]}],"products":[{"vendor":"universal-tool-calling-protocol","product":"typescript-utcp"}],"references":[{"url":"https://github.com/universal-tool-calling-protocol/typescript-utcp/security/advisories/GHSA-r8j5-8747-88cm","tags":["x_refsource_CONFIRM","web"]},{"url":"https://github.com/universal-tool-calling-protocol/python-utcp/issues/83","tags":["web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-45366","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-r8j5-8747-88cm"},{"url":"https://github.com/universal-tool-calling-protocol/typescript-utcp","tags":["package"]}],"sources":{"cve":[{"id":"CVE-2026-45366","title":"typescript-utcp: SSRF via attacker-controlled OpenAPI servers[0].url in HTTP communication protocol","description":"typescript-utcp is a typescript implementation of UTCP. Prior to 1.1.2, the @utcp/http package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. registerManual() validates the discovery URL against an HTTPS / loopback allowlist, but callTool() reuses the resolved toolCallTemplate.url directly without revalidating, and the OpenApiConverter 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: \"http://127.0.0.1:9090\" }] or servers: [{ url: \"http://169.254.169.254\" }]; the converter then produces tools whose URL points at internal services on the agent host. This vulnerability is fixed in 1.1.2.","status":"published","published":"2026-05-28T20:38:16.78Z","modified":"2026-05-29T13:56:10.659Z","cvss":[{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium","assessor":"cna"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T13:46:43.922571Z"}],"euvd":[{"id":"EUVD-2026-33053","description":"typescript-utcp is a typescript implementation of UTCP. Prior to 1.1.2, the @utcp/http package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. registerManual() validates the discovery URL against an HTTPS / loopback allowlist, but callTool() reuses the resolved toolCallTemplate.url directly without revalidating, and the OpenApiConverter 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: \"http://127.0.0.1:9090\" }] or servers: [{ url: \"http://169.254.169.254\" }]; the converter then produces tools whose URL points at internal services on the agent host. This vulnerability is fixed in 1.1.2.","status":"published","published":"2026-05-28T20:38:16Z","modified":"2026-05-29T13:56:10Z","cvss":[{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium"}],"updated_at":"2026-11-02T18:21:01.215325Z"}],"ghsa":[{"id":"GHSA-r8j5-8747-88cm","title":"@utcp/http: SSRF via attacker-controlled OpenAPI servers[0].url in HTTP communication protocol","description":"## Summary\n\nThe `@utcp/http` package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. `registerManual()` validates the discovery URL against an HTTPS / loopback allowlist, but `callTool()` reuses the resolved `toolCallTemplate.url` directly without revalidating, and the `OpenApiConverter` 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: \"http://127.0.0.1:9090\" }]` or `servers: [{ url: \"http://169.254.169.254\" }]`; the converter then produces tools whose URL points at internal services on the agent host.\n\nA separate prefix-bypass also affected the discovery-time check: the previous `startsWith('http://localhost')` guard let URLs like `http://localhost.evil.com` through.\n\n## Sister advisory\n\nThis is the npm/TypeScript counterpart of [GHSA-39j6-4867-gg4w / CVE-2026-44661](https://github.com/universal-tool-calling-protocol/python-utcp/security/advisories/GHSA-39j6-4867-gg4w) on the Python `utcp-http` package. Same vulnerability, same fix shape, same reporter.\n\n## Versions and patch state\n\n- `@utcp/http \u003c= 1.1.1` — vulnerable. Both the loopback-redirect (`http://127.0.0.1`) and the non-loopback internal-IP variants (e.g. `http://169.254.169.254`, `http://10.0.0.5`) succeed. Note: the `streamable_http` and `sse` `callToolStreaming` paths in 1.1.1 are TODO placeholders and don't actually fetch URLs, so the runtime SSRF surface in these protocols is currently confined to discovery — a future implementation must also call `ensureSecureUrl` before issuing the request.\n- `@utcp/http 1.1.2` — full fix. Runtime revalidation in `callTool` closes the non-loopback variants; the `OpenApiConverter` rejects, at conversion time, any spec fetched from a non-loopback source that declares a loopback `servers[0].url`, closing the loopback-redirect variant.\n\n## Impact\n\nA 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:\n- Map internal networks behind the agent.\n- Read AWS/GCP IAM credentials from cloud metadata endpoints (`http://169.254.169.254`, `http://metadata.google.internal`).\n- Reach unauthenticated internal services exposed on loopback (Elasticsearch, Redis HTTP, internal admin panels, the agent's own HTTP server).\n- Have responses returned to the LLM, which combined with prompt injection enables exfiltration back to the attacker.\n\n## Patch\n\nCommit on `dev`: 21f63e6.\n\nNew helper `packages/http/src/_security.ts` exposes `isSecureUrl`, `isLoopbackUrl`, `ensureSecureUrl`. Hostname-based validation closes the prefix bypass (`http://localhost.evil.com` → rejected). All three protocols' `registerManual` now call `ensureSecureUrl(url, 'manual discovery')`; `callTool` re-checks the resolved URL with `ensureSecureUrl(url, 'tool invocation')` immediately before the axios request. `OpenApiConverter` rejects remote spec → loopback server.\n\n## Workarounds\n\nFor users who cannot upgrade immediately:\n- Refuse to call `registerManual` with any URL controlled by an untrusted party, even over HTTPS.\n- Restrict outbound network access from the host running the agent so internal addresses (RFC1918, 169.254.0.0/16, loopback) are unreachable.\n\n## Credit\n\nDiscovered and reported by [@YLChen-007](https://github.com/YLChen-007) against the Python sibling implementation (universal-tool-calling-protocol/python-utcp#83). The TypeScript port shared the same code shape and the same vulnerability.","status":"published","published":"2026-05-14T20:55:05Z","modified":"2026-06-09T10:26:33Z","cvss":[{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T14:40:22.540562Z"}],"nvd":[{"id":"CVE-2026-45366","description":"typescript-utcp is a typescript implementation of UTCP. Prior to 1.1.2, the @utcp/http package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. registerManual() validates the discovery URL against an HTTPS / loopback allowlist, but callTool() reuses the resolved toolCallTemplate.url directly without revalidating, and the OpenApiConverter 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: \"http://127.0.0.1:9090\" }] or servers: [{ url: \"http://169.254.169.254\" }]; the converter then produces tools whose URL points at internal services on the agent host. This vulnerability is fixed in 1.1.2.","status":"published","published":"2026-05-28T22:17:00.773Z","modified":"2026-07-21T10:10:00.103Z","cvss":[{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium","assessor":"secondary"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T14:12:21.244426Z"}],"osv":[{"id":"GHSA-r8j5-8747-88cm","title":"@utcp/http: SSRF via attacker-controlled OpenAPI servers[0].url in HTTP communication protocol","description":"## Summary\n\nThe `@utcp/http` package is vulnerable to a blind Server-Side Request Forgery (SSRF) caused by a trust-boundary inconsistency between manual discovery and tool invocation. `registerManual()` validates the discovery URL against an HTTPS / loopback allowlist, but `callTool()` reuses the resolved `toolCallTemplate.url` directly without revalidating, and the `OpenApiConverter` 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: \"http://127.0.0.1:9090\" }]` or `servers: [{ url: \"http://169.254.169.254\" }]`; the converter then produces tools whose URL points at internal services on the agent host.\n\nA separate prefix-bypass also affected the discovery-time check: the previous `startsWith('http://localhost')` guard let URLs like `http://localhost.evil.com` through.\n\n## Sister advisory\n\nThis is the npm/TypeScript counterpart of [GHSA-39j6-4867-gg4w / CVE-2026-44661](https://github.com/universal-tool-calling-protocol/python-utcp/security/advisories/GHSA-39j6-4867-gg4w) on the Python `utcp-http` package. Same vulnerability, same fix shape, same reporter.\n\n## Versions and patch state\n\n- `@utcp/http \u003c= 1.1.1` — vulnerable. Both the loopback-redirect (`http://127.0.0.1`) and the non-loopback internal-IP variants (e.g. `http://169.254.169.254`, `http://10.0.0.5`) succeed. Note: the `streamable_http` and `sse` `callToolStreaming` paths in 1.1.1 are TODO placeholders and don't actually fetch URLs, so the runtime SSRF surface in these protocols is currently confined to discovery — a future implementation must also call `ensureSecureUrl` before issuing the request.\n- `@utcp/http 1.1.2` — full fix. Runtime revalidation in `callTool` closes the non-loopback variants; the `OpenApiConverter` rejects, at conversion time, any spec fetched from a non-loopback source that declares a loopback `servers[0].url`, closing the loopback-redirect variant.\n\n## Impact\n\nA 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:\n- Map internal networks behind the agent.\n- Read AWS/GCP IAM credentials from cloud metadata endpoints (`http://169.254.169.254`, `http://metadata.google.internal`).\n- Reach unauthenticated internal services exposed on loopback (Elasticsearch, Redis HTTP, internal admin panels, the agent's own HTTP server).\n- Have responses returned to the LLM, which combined with prompt injection enables exfiltration back to the attacker.\n\n## Patch\n\nCommit on `dev`: 21f63e6.\n\nNew helper `packages/http/src/_security.ts` exposes `isSecureUrl`, `isLoopbackUrl`, `ensureSecureUrl`. Hostname-based validation closes the prefix bypass (`http://localhost.evil.com` → rejected). All three protocols' `registerManual` now call `ensureSecureUrl(url, 'manual discovery')`; `callTool` re-checks the resolved URL with `ensureSecureUrl(url, 'tool invocation')` immediately before the axios request. `OpenApiConverter` rejects remote spec → loopback server.\n\n## Workarounds\n\nFor users who cannot upgrade immediately:\n- Refuse to call `registerManual` with any URL controlled by an untrusted party, even over HTTPS.\n- Restrict outbound network access from the host running the agent so internal addresses (RFC1918, 169.254.0.0/16, loopback) are unreachable.\n\n## Credit\n\nDiscovered and reported by [@YLChen-007](https://github.com/YLChen-007) against the Python sibling implementation (universal-tool-calling-protocol/python-utcp#83). The TypeScript port shared the same code shape and the same vulnerability.","status":"published","published":"2026-05-14T20:55:05Z","modified":"2026-09-10T03:50:47.363616586Z","cvss":[{"version":"3.1","score":4.7,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:L/A:N","level":"medium"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T15:46:55.480549Z"}]}}