docs: state the unofficial status, the real scan default, and the security route

.github/workflows/ci.yml

@@ -1,6 +1,6 @@
 # CI — supply-chain + secret + dependency-audit + build gates.
 #
-# Runs on push to main/refactor/v2 and every pull request.
+# Runs on push to main and every pull request.
 #
 # Layered security model:
 #   - Socket (supply-chain + SAST)  -> malicious-package / code-level gate
@@ -18,7 +18,7 @@ name: CI
 
 on:
   push:
-    branches: [main, refactor/v2]
+    branches: [main]
   pull_request:
 
 concurrency:

CONTRIBUTING.md

@@ -43,5 +43,5 @@ Prefixes, loosely [Conventional Commits](https://www.conventionalcommits.org/):
 
 ## Reporting security problems
 
-A private reporting channel is not set up yet. If you have found something, open an issue that says you have a
-security problem without the details, and a maintainer will get in touch to arrange a private conversation.
+Use the private route described in [SECURITY.md](SECURITY.md) rather than a public issue. A public issue that
+describes a vulnerability before it is fixed puts everyone running this tool at risk.

README.md

@@ -3,6 +3,8 @@
 A web interface for [Bumblebee](https://github.com/perplexityai/bumblebee), the open source supply chain
 security scanner from Perplexity AI.
 
+This is an unofficial interface for that scanner. It is not affiliated with or endorsed by Perplexity AI.
+
 Bumblebee walks a filesystem, reads package metadata across a dozen ecosystems, and reports anything that
 matches a published compromise. It is a CLI, which suits some people and not others. Bumblebee GUI wraps it in
 a web app: point it at a directory, run a scan, watch progress stream in, and read the result as a table you
@@ -175,8 +177,8 @@ and removes the partial output.
 | `deep` | Incident response | Explicit root paths, full home directory |
 
 `deep` is the incident-response profile and refuses to run without an explicit root, because it will not guess
-which paths matter. The scan form's max duration defaults to `10m`, which a deep scan of a whole home
-directory can exceed.
+which paths matter. The scan form leaves max duration empty by default, which sends no limit at all and keeps
+the scanner's own behaviour. Set `30s`, `10m` or `2h` to cap a scan, or `0` for no limit.
 
 ## What has actually been tested
 

SECURITY.md

@@ -0,0 +1,42 @@
+# Security policy
+
+## Reporting a vulnerability
+
+Use GitHub's private vulnerability reporting for this repository: open the **Security** tab and choose **Report a vulnerability**. The report stays private between you and the maintainer until a fix is published.
+
+Please do not open a public issue for a vulnerability.
+
+A useful report says:
+
+- what the problem is, and where in the project it lives
+- how to reproduce it, including the commit or image you tested
+- what an attacker could do with it
+- a suggested fix, if you have one
+
+There is no bug bounty and no guaranteed response time. This is a small project maintained in spare time. Reports are read and answered as soon as possible.
+
+## Supported versions
+
+Only the latest commit on `main` is supported. There are no maintained release branches yet.
+
+## Scope
+
+In scope:
+
+- the backend API and the frontend in this repository
+- the Docker Compose setup and its default configuration
+- the image builds and the bundled exposure catalogues
+
+Out of scope:
+
+- the upstream Bumblebee scanner and its catalogues. Report those to
+  [perplexityai/bumblebee](https://github.com/perplexityai/bumblebee/issues).
+- a vulnerability in a dependency with no exploitable path through this project
+
+## How this project runs
+
+Bumblebee GUI is an unofficial interface to the Bumblebee scanner, not affiliated with Perplexity. It runs on the host that starts it. The API binds to loopback by default, and a directory is only scanned if it is mounted into the backend container, read-only, by whoever starts the stack. Nothing is sent anywhere except the upstream threat-intel downloads performed during the image build.
+
+## Reporting something that is not a vulnerability
+
+Misleading output is worth reporting as a normal issue: a scan that matches nothing and a scan whose catalogues are stale should not look alike, and a scan that stopped early should say so. If you find a case where the interface states more confidence than the data supports, open an issue.