Firebase APK Scanner runs a bundled shell script on one APK or a folder of them. The script decompiles each app with apktool, pulls the Firebase settings out of every place they can hide, and tests the backend for misconfigurations: authentication, Realtime Database, Firestore, Storage buckets, Cloud Functions and Remote Config.
The agent then reads the JSON and text reports and presents a summary, the extracted configuration, vulnerabilities with severity, and fixes. It is meant for authorized security research only, and it is set to run only when you invoke it.
What You Get
- A scan summary with counts of vulnerable, failed and untested APKs
- The extracted Firebase configuration: project ID, database URL, storage bucket, API key and auth domain
- Findings graded critical, high, medium and low with brief evidence
- Remediation advice for each finding, backed by a vulnerability patterns reference
- A manual testing fallback for cases where the scanner writes no report
Common Use Cases
- Mobile app security assessments of Firebase-backed Android apps
- Checking an app you own for open database rules before release
- Testing for open signup, anonymous auth and email enumeration
- Enumerating Cloud Functions for unauthenticated access in an authorized penetration test
What Makes It Unique
Most mobile audit skills stop at reading configuration. This one goes on to test the live Firebase endpoints, and it reports what it could not test so that a failed or config-less APK is never shown as a clean result. It states its limits up front: Cloud Functions are probed only in us-central1, and auth, functions and Remote Config are tested only for the first project ID and key found. It also carries a list of rationalizations to reject, such as “the API key is public anyway”.
Under The Hood
The skill runs a 50 KB bundled shell script, scanner.sh, and is allowed to use apktool and curl. The script decompiles APKs, then sends requests over the network to the Firebase endpoints it extracted: it reads and writes test entries in databases and storage and removes them afterwards, tries signups, and probes functions and Remote Config. It writes report files to a results folder in the current directory. The skill is marked so that the model does not start it on its own, and it must only be pointed at apps you are authorized to test.