Skip to content
Android R8 Keep-Rule Analyzer

Android R8 Keep-Rule Analyzer

Apache-2.0
Repository Docs
markdown Development
androidr8proguardapp-sizegradle

Summary

Google's official Android skill for auditing R8 and ProGuard keep rules — finds redundant and overly broad rules that quietly inflate APK size.

Features

  • Detects redundant, overly broad and library-subsuming keep rules
  • Checks AGP version and R8 full-mode configuration first
  • Quantitative path on R8 9.3.7-dev+, heuristic fallback on older versions
  • Installable via the Android CLI: android skills add --skill=r8-analyzer

Install This Skill

Add this skill to your favorite AI agent in a few steps.

Any AI agent

This skill is plain instructions — it works with any assistant that accepts custom instructions or system prompts.

  1. Copy the skill content with the button below.
  2. Paste it into your agent's instruction file or system prompt (for example AGENTS.md, .cursorrules, or a custom instructions field).
  3. Ask the agent to apply the skill whenever the task matches.

Skill Content

Markdown Content

Copy this content and use it with your preferred AI agent

---
name: r8-analyzer
description: Analyzes Android build files and R8 keep rules to identify redundancies,
  broad package-wide rules, and rules that subsume library consumer keep rules. Use
  when developers want to optimize their app's size, remove redundant or overly broad
  keep rules, or troubleshoot Proguard configurations.
license: Complete terms in LICENSE.txt
metadata:
  author: Google LLC
  last-updated: '2026-07-18'
  keywords:
  - R8
  - proguard
  - keep rules
  - app size
  - optimization
---

## Step 1. Setup and configuration check

- Inspect `build.gradle`, `build.gradle.kts`, and `gradle.properties`.
- Use [references/CONFIGURATION.md](references/CONFIGURATION.md) to identify missing optimizations.
- **AGP** : If \< 9.0, suggest migration to 9.0 for [build time improvement
  performance](references/android/topic/performance/app-optimization/enable-app-optimization.md)
- **Full Mode** : Verify `android.enableR8.fullMode=false` is removed from gradle.properties.

## Step 2. Analysis path selection

- Inspect `build.gradle`, `build.gradle.kts`, and `gradle.properties` and
  `libs.versions.toml` to get the R8 version

- **If R8 \>= 9.3.7-dev** : Proceed to **Path A (Quantitative)**.

- **If R8 \< 9.3.7-dev** : Proceed to **Path B (Heuristic)**.

### Path A: Quantitative data generation (R8 \>= 9.3.7-dev)

- **Check requirements** : Python and `protobuf` package are mandatory.
- **Generate and analyze** : You MUST run the shell commands described in [references/CONFIGURATION-ANALYZER.md](references/CONFIGURATION-ANALYZER.md) to generate the proto file using R8 configuration analyzer, convert it to json and analyze the result.
- **Report** : Rely entirely on the generated file `analysis.txt` for scores and rule impact metrics. Proceed to Step 3.

### Path B: Heuristic evaluation and recommendation (R8 \< 9.3.7-dev)

*(Use ONLY if quantitative data generation is not possible)*

- **Manual evaluation** : Inspect `proguard-rules.pro`.
- **Library check** : Compare rules against [references/REDUNDANT-RULES.md](references/REDUNDANT-RULES.md). Suggest **Remove** for bundled rules.
- **Custom rule check** : Use [references/KEEP-RULES-IMPACT-HIERARCHY.md](references/KEEP-RULES-IMPACT-HIERARCHY.md) and [references/REFLECTION-GUIDE.md](references/REFLECTION-GUIDE.md) to prioritize and evaluate. Suggest **Refine** for broad rules (for example, package-wide).
- **Validation** : Suggest Macrobenchmark tests using [UI Automator](references/android/training/testing/other-components/ui-automator.md) for any proposed changes. Proceed to Step 3.

## Step 3. Report generation

- **Format** : Follow [references/REPORT_FORMAT.md](references/REPORT_FORMAT.md) strictly.
- **Input**: Extract metrics (Scores, Impacts, Example Classes) directly from generated file analysis.txt if using Path A, or from manual findings if using Path B.
- **Output** : Output ONLY the raw Markdown report in the chat. Do NOT output conversational filler (for example, "Here is your report..."). Do NOT provide recommendations, next steps, or any other text outside of the sections defined in [references/REPORT_FORMAT.md](references/REPORT_FORMAT.md) Do NOT mention the path used for analysis of the configuration

## Constraints

- **Strict output limit**: The final output MUST strictly be the Markdown report and nothing else.
- **No code changes**: Research and suggest only; Do not modify files.
- **No redundancy**: Do not explain R8 benefits or reference skill internal files in the report.
- **Focus**: Omit sections (for example, Subsumed Rules, Configuration) if no issues or items are found.

Description

Keep rules accumulate. A -keep class com.example.** added years ago to fix one crash now pins an entire package out of shrinking, and nothing in the build output tells you. This is Google's own Android skill for finding those rules and quantifying what they cost.

It starts by reading build.gradle, build.gradle.kts, gradle.properties and libs.versions.toml to establish the configuration — flagging an AGP below 9.0, and checking whether android.enableR8.fullMode=false is still suppressing full-mode optimisation. It then branches on the R8 version: on 9.3.7-dev and later it takes a quantitative path using R8's own generated data, and on older versions it falls back to a heuristic analysis.

What it looks for: rules that duplicate each other, rules broad enough to subsume the consumer keep rules a library already ships, and package-wide wildcards where a narrower rule would do. The output is aimed at a decision — which rules to remove and what that is worth in size — rather than a list of warnings.

Part of Android's official skills repository, which deliberately targets workflows where evaluations show LLMs underperform rather than well-trodden ground like Compose basics. Apache-2.0 licensed and installable through the Android CLI (android skills add --skill=r8-analyzer) or directly into any agent that reads SKILL.md.

Related Skills

New

GreenSock's official ScrollTrigger skill — scroll-linked animation, pinning, scrub and trigger positioning, with the exact start/end syntax agents get wrong.

DevelopmentDesign & Creative

Microsoft's official skill that turns an agent into an Azure solution architect — 10 design principles, 6 architecture styles, 44 design patterns and a Well-Architected review workflow.

Development

Trail of Bits' security review skill for PRs, commits, and diffs: risk-first analysis with git history, blast radius, and honest coverage limits.

Development

Skill: CAD

by earthtojake

New

Generates and validates parametric STEP-first CAD parts and assemblies from natural language.

Development
Browse all skills →