Registry HTTP actions could reach private network destinations before nsjail egress filtering
6.5
Medium
CVSS 3.1 base
What to do
If you run Tracecat 0.41.0 to before 1.0.0-beta.52-rc.3, upgrade to 1.0.0-beta.52-rc.3 or later.
Anyone who can reach it over the network, with a low-privilege account and no user interaction, can read sensitive data.Read from the CVSS 3.1 vector
To exploit it
What it gets
My analysis of this finding: root cause, the vulnerable code, the impact and the fix, in the site’s format. The official advisory of record is linked above.
| Package | Ecosystem | Vulnerable | Fixed in |
|---|---|---|---|
Tracecat | other | >= 0.41.0, < 1.0.0-beta.52-rc.3 | >= 1.0.0-beta.52-rc.3 |
Tracecat’s core.cases.upload_attachment_from_url registry action takes a URL from an authenticated workflow author, fetches it with httpx, and stores the full response body as a case attachment the author can read back. There is no URL validation of any kind: no scheme allowlist, no private or link local IP blocking, no DNS rebinding guard. That makes it a clean full read SSRF, where the response is handed straight back to the attacker rather than merely triggered blind.
The action is defined in packages/tracecat-registry/tracecat_registry/core/cases.py. The url parameter is a plain str, and the fetch runs with no checks between input and request:
async def upload_attachment_from_url( case_id: Annotated[str, Doc("The ID of the case to attach the file to.")], url: Annotated[str, Doc("The URL of the file to upload.")], headers: Annotated[dict[str, str] | None, Doc("The headers to use when downloading the file.")] = None, file_name: Annotated[str | None, Doc("...")] = None,) -> types.CaseAttachmentRead: """Upload a file attachment to a case from a URL.""" async with httpx.AsyncClient() as client: response = await client.get(url, headers=headers) response.raise_for_status() content = response.content content_type = response.headers.get("Content-Type") ... return await _upload_attachment(case_id, file_name, content, content_type)The bugfetches the author supplied url with no validationThe bugstores the full response body as a readable attachment
Two details sharpen it. The headers parameter is attacker controlled, so a request can carry the X-aws-ec2-metadata-token header needed for IMDSv2 token based metadata access. And content is persisted through _upload_attachment, so whatever the internal service returns becomes a downloadable attachment on the case. The same gap in reflected form sits in the core.http_request family in core/http.py, where Pydantic HttpUrl validates URL shape but never rejects private ranges.
My own findings pair this with Tracecat’s default of TRACECAT__DISABLE_NSJAIL=true, which ships across .env.example, both executor services in docker-compose.yml, and tracecat/config.py. With the sandbox off the executor has no network isolation, so the SSRF reaches everything on the Docker network. I live validated reachability from the executor to Redis (received +PONG), PostgreSQL (TCP connect on 5432), and MinIO (HTTP 200). On a cloud deployment the same action reaches 169.254.169.254 for IAM credential theft, with the custom headers covering IMDSv2. Because the full body is stored and retrievable, this is exfiltration, not a pingback. One mitigating note: httpx does not follow redirects by default, so redirect based pivots do not work out of the box.
Validate before the request, not just the format. Resolve the hostname, reject any resolved address inside 127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16, 169.254.0.0/16, ::1/128, fc00::/7 and fe80::/10, and restrict the scheme to http and https. Apply that same check in upload_attachment_from_url and across the core.http_request and core.http_paginate actions, since the paginator can emit arbitrary follow up URLs that skip even the HttpUrl format check.
Sanaan Fayaz Wani (GitHub sfwani) reported this vulnerability to the Tracecat maintainers under coordinated disclosure and is credited as a reporter in GHSA-8p4j-2mm9-rh78, published 2026-09-20.
All published findings: advisory index.