How to use the Android Permissions Security Auditor agent skill

What the skill makes the agent do

Once the skill is installed, the agent knows what to look for when you ask it to review an Android project. It checks for weak custom permissions, overly broad file or URI access, caller identity checks that can be spoofed, and poor permission request flows. For each problem it finds, it suggests a specific fix.

The skill works from reference notes and guidance written into it, so the advice follows the documented Android approach rather than the agent's general memory. The playbook covers areas such as weak custom permission protection, overly broad URI permission grants, missing ContentProvider split permissions, and insecure broadcasts.

• Weak custom permissions: suggests moving to signature or signature with knownSigner.
• Broad URI grants: flags a wildcard pathPrefix of "/" and suggests a scoped subpath.
• Spoofable caller checks: flags string extras such as calling_package and suggests Binder.getCallingUid().
• Runtime permission flows: suggests a three-state flow and handling of permanent denial.
• Storage access: suggests migrating to the Photo Picker where it applies.

Install the skill in Claude

You can install the skill in one of two ways. Download the ZIP from this page and unzip it into your personal skills folder, ~/.claude/skills/, so it is available in every project. Or unzip it into a project's .claude/skills/ folder so it applies only to that project and can be shared with your team through the repository.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload the skill

Check that your project meets the requirements

The skill needs an Android project to inspect, including the manifest and source files. The suggested code relies on AndroidX dependencies, so add androidx.core:core-ktx and androidx.activity:activity-ktx (or androidx.fragment:fragment-ktx) to the app-level build.gradle file. Some features depend on your SDK levels.

• An Android project with AndroidManifest.xml and source files to inspect.
• AndroidX dependencies core-ktx and activity-ktx (or fragment-ktx) for the suggested code.
• compileSdk and targetSdk of 31 or higher if you need knownSigner partner permissions.
• compileSdk and targetSdk of 34 or higher if you need broadcast identity sharing with BroadcastOptions.setShareIdentityEnabled.

Trigger the skill with a natural request

You do not need a special command. Describe the task in plain language and mention the files or areas you care about. The skill's description covers reviewing AndroidManifest.xml, defining custom permissions, securing Services, Receivers and Providers, verifying caller identity, and implementing runtime permission flows, so requests that name these areas are most likely to load it.

Be specific about scope. A request to review one manifest or one provider gives more useful results than a request to check an entire large codebase at once.

• "Review my AndroidManifest.xml for exported components and weak custom permissions before release."
• "Check my ContentProvider for overly broad URI permission grants and missing read and write permissions."
• "Audit the bound service in this project for caller identity checks I can spoof."
• "Review how my app requests location and storage permissions and suggest a better flow."

Read and act on the results

Treat the output as a list of findings to verify, not a final verdict. For each finding, check the cited file and line against your code, confirm the risk applies to your app's exported components and callers, and then apply or adapt the suggested fix. Re-run the review after changes to confirm the issue is gone.

The skill is built from reference material on the Android permission model, including the permission element documentation, the permissions overview, guidance on sender of pending intents, and the runtime permission request training. Where a finding points to one of these topics, the reference notes give the underlying rule.

• Verify each finding against the actual code path.
• Apply the suggested fix, then rebuild and test the affected flows.
• Ask the agent to re-check the same files after you make changes.

Safety notes: scripts and network access

This skill does not run scripts and does not use the network, according to its listing. It provides guidance for the agent to apply to the files you share with it. Review its suggested changes yourself before committing them, since the skill can miss issues or misread code without human review.

• Runs scripts: no
• Uses network: no
• The skill gives guidance, not a verified audit. Keep a human reviewer in the loop for release decisions.