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 agentmods add skills/expensify/app/agent-device-evidencenpx skills add Expensify/App --skill agent-device-evidencegit clone --depth 1 https://github.com/Expensify/AppWrote 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/expensify/app/agent-device-evidence)<a href="https://agentmods.dev/skills/expensify/app/agent-device-evidence"><img src="https://agentmods.dev/badge/skills/expensify/app/agent-device-evidence.svg" alt="Measured on agentmods" 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 | $0.00043 | $0.03491 |
| Opus 5 | $0.00022 | $0.01746 |
| Sonnet 5 | $0.00009 | $0.00698 |
| Haiku 4.5 | $0.00004 | $0.00349 |
Grade A, and why
agent-device-evidence 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 3d 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 — 237 lines — stays where its author put it; the contents beside it link to each section on GitHub.
agent-device-evidence
Records iOS: Native and Android: Native MP4 evidence for the test or repro steps declared in an Expensify GitHub PR or issue. The source of truth is the test/repro steps themselves, not the surrounding code or context - the skill works equally well on a PR's ### Tests section, an issue's ## Action Performed: block, or any future Markdown body where steps are clearly authored.
Specializes the agent-device skill: delegates device lifecycle (bundle ID, Metro, device pick, session, open) to its Bring-up, then captures one artifact per declared flow per platform, writes a JSON manifest, and surfaces local file paths.
The skill is autonomous and non-interactive. It never pauses for user input mid-run. All inputs are provided at invocation time; all failures surface as structured errors with exit codes.
HybridApp-only (the parent skill's pre-flight enforces this). Standalone (non-HybridApp) builds are out of scope - production mobile evidence runs against HybridApp.
Scope
In scope: iOS: Native (iOS Simulator), Android: Native (Android Emulator), HybridApp dev build only. Inputs may come from PRs or issues - the skill does not gate on code changes.
Out of scope: Android: mWeb Chrome, iOS: mWeb Safari, iOS: mWeb Chrome, Windows: Chrome, MacOS: Chrome / Safari. Decline with EXIT 4 and point to a browser-driver skill (playwright-app-testing). Standalone (non-HybridApp) builds. Decline with EXIT 7 BRING_UP_FAILED per the parent skill's gate.
Inputs
| Input | Source | Required |
|---|---|---|
| Source URL (PR or issue) | First positional arg, e.g. https://github.com/Expensify/App/pull/89475 or .../issues/89855 |
Yes |
--platforms ios,android |
Flag | No (default: derived) |
-e KEY=VALUE step-param overrides |
Repeatable | No |
Bare numbers are rejected (PRs and issues share the GitHub number namespace; the URL path is the safe disambiguator). No interactive prompts.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 237 lines · 43 tokens per session scan A 850193f720a1
agent-device-evidence is a skill published in the GitHub repository Expensify/App (5,017 stars, last pushed yesterday), licensed MIT. It adds 43 tokens to every session and 3,491 once invoked, about $0.0002 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 skills, from other repositories
bump-android-version
Use when bumping Maestro's Android compileSdk/targetSdk to a new API level and validating end-to-end against the test-e2e GHA workflow until the test-android job is green.
solopi-ai
通过 SoloPi 的机器可读 CLI 编译和执行 AI 验证计划，管理签名端侧 ExecuTorch 决策模型、持久设备池、无人值守任务、安卓设备、应用、动作、配置、用例步骤与交互录制、回放及性能历史、动态 Agent、批量与重复执行、性能监控、压力测试和证据。适用于需求/AC 到 Result Judge 三态结论、cloud/on-device 决策切换、模型发布门禁，以及 generation 租约的多设备 CI 执行。.
dogfood
Systematically explore and test a mobile app on iOS/Android with agent-device to find bugs, UX issues, and other problems. Use when asked to dogfood, QA, exploratory test, find issues, bug hunt, or test this app on mobile.
argent-test-ui-flow
Autonomously test an app UI (iOS or Android) by running interact-screenshot-verify loops using argent MCP tools. Use when testing UI flows, verifying login works, testing navigation, running end-to-end UI test scenarios, manual QA steps, visible UI changes, or visual behavior.
flutter-add-integration-test
Configures Flutter Driver for app interaction and converts MCP actions into permanent integration tests. Use when adding integration testing to a project, exploring UI components via MCP, or automating user flows with the integrationtest package.
mobile-automation
Control Android and iOS devices, emulators and simulators — launch apps, tap, swipe, type, take screenshots, read the accessibility tree. Use when a task involves a mobile device or app, mobile UI testing, or reproducing a bug on a phone.