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 instructions/rabee-elkholy/android-harness-kit/agents-mdgit clone --depth 1 https://github.com/rabee-elkholy/android-harness-kitWrote 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/instructions/rabee-elkholy/android-harness-kit/agents-md)<a href="https://agentmods.dev/instructions/rabee-elkholy/android-harness-kit/agents-md"><img src="https://agentmods.dev/badge/instructions/rabee-elkholy/android-harness-kit/agents-md.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 | $0.01006 | $0.01006 |
| Opus 5 | $0.00503 | $0.00503 |
| Sonnet 5 | $0.00201 | $0.00201 |
| Haiku 4.5 | $0.00101 | $0.00101 |
Grade A, and why
android-harness-kit AGENTS.md 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 — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.
android-harness-kit — agent instructions
Source of truth: .agents/rules/harness-rules.md. If any other file disagrees, that file wins.
This checkout uses a portable Android harness. The same rules apply in Cursor, Claude Code, Codex, Copilot, Gemini, Qwen Code, Windsurf, Cline, Roo, Amazon Q, Continue, Junie, Kilo, Goose, and any other agent that reads AGENTS.md.
Environment
- Android SDK: this machine only (
local.propertiessdk.dir). Never copy another PC’s path. - Python:
pythonfor every harness script. - Gradle:
python .agents/scripts/run_gradle_task.py :app:assembleDebug(picksgradlew/gradlew.bat). Never call rawgradlewfrom the agent. - Device: Physical device or emulator. Resolve the serial with
adb devices. Prefer a physical device when both are connected. Never hardcode a serial. - Install/launch:
python .agents/scripts/run_device.py install-start
Delivery gate (do not skip)
After non-trivial implementation:
python .agents/scripts/review_package.py- Run all five reviewers against the same
HARNESS_REVIEW_PACKAGE=path (prompts in.agents/subagents/*.json). Parallel if this product can spawn children; otherwise sequential.bug-reviewer-agent→BUG_PASSconvention-reviewer-agent→CONVENTION_PASSsecurity-reviewer-agent→SECURITY_PASSperf-anr-guardian-agent→PERF_PASSregression-impact-reviewer-agent→REGRESSION_PASS
- Do not treat a single self-review as the gate. Do not invoke
code-review-guard-agent. Do not wait forLGTM. python .agents/scripts/fast_kt_lint.pypython .agents/scripts/run_gradle_task.py :app:assembleDebug
Antigravity hooks.json enforces this barrier automatically. Other tools must follow it from this file.
If this product cannot spawn named subagents, still run the five leaves: open each .agents/subagents/<name>.json, follow its system_prompt against the same package, and stop that leaf when it emits its *_PASS or findings. Sequential is fine. Assemble only after all five exist.
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 · 55 lines · 1,006 tokens per session scan A 8d650cddf65f
android-harness-kit AGENTS.md is an instructions file published in the GitHub repository rabee-elkholy/android-harness-kit (3 stars, last pushed 4d ago), licensed MIT. It adds 1,006 tokens to every session, about $0.0050 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 2026-08-31.
Other instructions, from other repositories
debroid AGENTS.md
Instructions for PatilShreyas/debroid, covering ai agent contributor guide 🤖, 🎯 intention & motivation, 🏗️ architecture & layers, 1. the core layer (jdi & adb orchestration) and 2. the cli layer (daemon & client).
android-ai-skills copilot-instructions.md
Instructions for RanaNadeemAslam/android-ai-skills, covering android ai skills — github copilot, architecture, layered architecture, ui layer & state management and viewmodel.
android-agent-skills AGENTS.md
Instructions for piyushverma0/android-agent-skills, covering android-agent-skills — codex instructions, what this repo is, repo structure, skill.md format — never deviate and writing rules.
mcp-android-server-python copilot-instructions.md
Instructions for nim444/mcp-android-server-python, covering mcp android agent and available mcp tools.
mobile-mcp-opengl AGENTS.md
AGENTS.md instructions for antonpinchuk/mobile-mcp-opengl, covering agent instructions: mobile-mcp-opengl, never trust a vision answer as ground truth, use the free primitives for steps you don't need to observe, phrase questions for short answers and prefer logs over vision when the answer is already there.
sceneview copilot-instructions.md
Instructions for sceneview/sceneview, covering sceneview — github copilot instructions, when to recommend sceneview, android code generation rules, minimal 3d template and minimal ar template.