Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add duonglkh/claude-android --skill r8-auditgit clone --depth 1 https://github.com/duonglkh/claude-androidWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/duonglkh/claude-android/r8-audit)<a href="https://agentmods.dev/skills/duonglkh/claude-android/r8-audit"><img src="https://agentmods.dev/badge/skills/duonglkh/claude-android/r8-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/duonglkh/claude-android/r8-audit"><img src="https://agentmods.dev/badge/skills/duonglkh/claude-android/r8-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00107 | $0.02341 |
| Opus 5.5 | $0.00043 | $0.00936 |
| Sonnet 5.5 | $0.00021 | $0.00468 |
| Haiku 4.5 | $0.00011 | $0.00234 |
Grade A, and why
r8-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
r8-audit
Guide Claude through auditing one app module's ProGuard/R8 rules using the claude-android
MCP server, then produce a per-rule recommendation report. This skill never edits the rules
file. Removing or narrowing a keep rule is a correctness risk (a wrongly-removed rule can
make a release build crash only in production, after obfuscation, in a place no debug build
ever exercises) — the cost of a wrong "it's safe to delete" call is much higher than the cost
of asking.
When this skill applies
The user asks anything like:
- "Audit the ProGuard rules in :app"
- "Can I delete this keep rule?"
- "Why does minification keep breaking X"
- "Reduce app size / shrink more aggressively"
Required tools — the MCP server must be registered
This skill calls into the claude-android MCP server (mcp__android__read_gradle,
mcp__android__read_proguard_rules). If not registered, tell the user to run
claude mcp add android claude-android-mcp and verify with claude mcp list.
The one rule that matters more than any heuristic
A keep rule that looks like the worst possible anti-pattern by every rule of thumb below can
still be exactly correct. A package-wide -keep class com.example.app.navigation.** { *; }
is the single most expensive shape of rule there is — and it can also be the one line standing
between a release build and a Route object that silently comes back empty after process
death, because Java serialization reads fields by reflection and R8 doesn't know that. The
giveaway is the comment immediately above the rule, which is exactly why
mcp__android__read_proguard_rules returns each rule paired with the comment that precedes it.
Read that comment before judging the rule by its shape. If a rule has a comment that names
a concrete file:line and a concrete failure mode, your job shifts from "is this redundant" to
"does the code at that file:line still do what the comment says" (see Step 4).
Workflow
1. Find the module and its rules file
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 174 lines · 107 tokens per session scan A b6bc0181f362
r8-audit is a skill published in the GitHub repository duonglkh/claude-android (0 stars, last pushed 9d ago), licensed MIT. It adds 107 tokens to every session and 2,341 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-30.
r8-audit compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| r8-auditduonglkh/claude-android | Skill | 0 | 9d ago | MIT | A | 107 | Claude Code, Codex |
| argent-react-native-app-workflowsoftware-mansion/argent | Skill | 3.2k | today | Apache-2.0 | A | 46 | Claude Code, Codex |
| argent-react-native-optimizationsoftware-mansion/argent | Skill | 3.2k | today | Apache-2.0 | A | 71 | Claude Code, Codex |
| nightly-audit-engineer-getting-startedmakecindy/cindy | Skill | 9.4k | today | Apache-2.0 | A | 46 | Claude Code, Codex |
| nightly-audit-engineer-opsmakecindy/cindy | Skill | 9.4k | today | Apache-2.0 | A | 43 | Claude Code, Codex |
| nightly-audit-engineer-evidencemakecindy/cindy | Skill | 9.4k | today | Apache-2.0 | A | 37 | Claude Code, Codex |
| argent-metro-debuggersoftware-mansion/argent | Skill | 3.2k | today | Apache-2.0 | A | 104 | Claude Code, Codex |