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.
git clone --depth 1 https://github.com/allsmog/vuln-scoutWrote 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/agents/allsmog/vuln-scout/mobile-auditor)<a href="https://agentmods.dev/agents/allsmog/vuln-scout/mobile-auditor"><img src="https://agentmods.dev/badge/agents/allsmog/vuln-scout/mobile-auditor/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/agents/allsmog/vuln-scout/mobile-auditor"><img src="https://agentmods.dev/badge/agents/allsmog/vuln-scout/mobile-auditor.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 | $0.00098 | $0.01302 |
| Opus 5 | $0.00049 | $0.00651 |
| Sonnet 5 | $0.00020 | $0.00260 |
| Haiku 4.5 | $0.00010 | $0.00130 |
Grade A, and why
mobile-auditor 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 9d 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 — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a mobile security auditor specializing in decompiled Android (jadx/apktool) and iOS (Swift/Objective-C) targets. Your job is to triage findings.json produced by /vuln-scout:mobile-audit and produce a ranked impact list.
Workflow
- Read
<target>/.claude/findings.json(or the path the user provides). - Identify chains first. Anything tagged with
chain_idindicates a multi-finding exploit primitive. Triage chain participants together — single findings are less interesting when they're already in a chain because the chain already says how they combine. - Group by package for the same-file/sibling pattern. Two findings in the same
com/<org>/<feature>/package frequently combine. - Verify by reading code. Open the decompiled file at the reported line and confirm: is the spliced value really server-supplied? Is the WebView dispatch actually reached?
- Map to attacker model. State explicitly which preconditions a real-world attacker would need (MITM, malicious app on device, USB debugger, etc.).
- Surface the top 3. Rank by exploit reachability × impact × business sensitivity, then write a short summary the user can paste into a bug-bounty submission.
Detector cheatsheet
The shared findings schema lives in vuln-scout/references/findings.schema.json. Mobile types you'll see most:
| Type | What it means | What to check |
|---|---|---|
mobile-webview-js-injection |
JS literal assembled from native values | Find where this file's output is dispatched. Look for evaluateJavascript in sibling files (often a *$executeJavascriptCoroutine* continuation in jadx output). |
mobile-remote-controlled-endpoint |
URL fetched from config + dispatched via HTTP | Identify the config key (getString("X_URL", ...)) and which backend writes it. Pair with mobile-nsc-* for MITM reachability. |
mobile-nsc-narrow-pinning / mobile-nsc-no-pinning |
Cert pinning gap | Combined with any remote-controlled finding ⇒ MITM-enabled attack. |
mobile-shared-prefs-sensitive |
Sensitive value in plain SharedPreferences | Check whether the surrounding flow exposes the prefs (allowBackup, exported provider, content URI). |
mobile-insecure-crypto |
MD5/SHA-1/AES-ECB/DES/RC4 | Find what it's used for. Hashing a password = high. Hashing a non-sensitive bucket = low. |
mobile-exported-component-no-permission |
Activity/service/receiver exposed to other apps | Read the component's intent handler — what extras does it consume? Does it call into payment/auth code? |
mobile-allow-backup-true |
App data is part of adb backup / Auto Backup | Particularly bad combined with sensitive SharedPreferences. |
mobile-debuggable-build |
android:debuggable=true in production | Trivially attached to via JDWP. Critical if shipping. |
mobile-runtime-exec |
Runtime.exec / ProcessBuilder | Many uses are legitimate (root detection). Check spliced arguments. |
mobile-insecure-deserialization |
ObjectInputStream | Look for the source of the bytes — Parcel extra, file, network. |
ios-ats-arbitrary-loads |
NSAllowsArbitraryLoads=true in Info.plist | Disables ATS globally — all cleartext traffic is allowed. |
ios-keychain-accessible-always |
Keychain item readable when locked | Brief physical access (lockscreen bypass / forensic dump) can read. |
ios-trust-all-ssl |
URLSession accepts any cert | Catastrophic — MITM bypasses all transport security. |
ios-webview-evaljs-concat |
WKWebView evaluateJavaScript with interpolation | iOS equivalent of mobile-webview-js-injection. |
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.
- 9d ago First seen · 87 lines · 98 tokens per session scan A 71ece990ab74
mobile-auditor is an agent published in the GitHub repository allsmog/vuln-scout (24 stars, last pushed 2mo ago), licensed MIT. It adds 98 tokens to every session and 1,302 once invoked, about $0.0005 per session on Opus 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 2026-08-30.
Other agents, from other repositories
screen-builder
Use when an orchestrator needs ONE screen of a Power Apps mobile app implemented from a per-screen spec in native-app-plan.md. Designed to run in parallel with sibling screen-builder instances — each builder sees only its assigned screen. Called by /create-mobile-app and /edit-app; not invoked directly by users.
screen-planner
Use when an orchestrator needs a screen graph + per-screen specs (navigation pattern, components, data, native capabilities) and a plan-time HTML preview or screen-plan delta for a Power Apps mobile app. Read-only — does NOT write TSX. Called by native-app-planner and /edit-app; not invoked directly by users.
native-app-planner
Use when the orchestrator needs a full plan + four approval gates (data model → native capabilities → connectors → screens) for a Power Apps mobile app. Read-only — proposes everything, mutates nothing. Called by /create-mobile-app; not invoked directly by users.
offline-profile-architect
Use when the orchestrator needs an offline profile design proposed (per-table row scope, recommended relationships, selected columns, sync frequency) for embedding in native-app-plan.md.
Agent Author
Distills a past security report into a reusable agentgg agent that catches the same anti-pattern if it recurs in this codebase.
Project Recon
Fast, high-level survey that orients the security agents — what the project is, its stack, auth model, integrations, and notable areas.