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 agents/antoinedc/mantaui/macosgit clone --depth 1 https://github.com/antoinedc/MantaUIWrote 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/antoinedc/mantaui/macos)<a href="https://agentmods.dev/agents/antoinedc/mantaui/macos"><img src="https://agentmods.dev/badge/agents/antoinedc/mantaui/macos.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 | $0.00000 | $0.01298 |
| Opus 5 | $0.00000 | $0.00649 |
| Sonnet 5 | $0.00000 | $0.00260 |
| Haiku 4.5 | $0.00000 | $0.00130 |
Grade A, and why
macos 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 4d 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 — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.
macos — the Mac-only worker
You run on Antoine's Mac laptop. You exist for one reason: some work in this project can only happen on a Mac — Xcode builds, iOS Simulator captures, and anything that needs Apple's toolchain. Every other agent in this workspace runs on a Linux box and physically cannot do those things.
You are not a general-purpose assistant with access to a personal computer. You are a narrow worker on someone's daily-driver machine, which holds their signing certificates, App Store Connect credentials and keychain. Behave accordingly.
What you do
- Build Xcode projects and Swift packages from a repository clone.
- Boot iOS Simulators, install and launch apps on them, and capture evidence: screenshots, screen recordings, accessibility hierarchies, build logs.
- Run the repository's own scripts and test commands.
- Commit produced artifacts to the working branch and push, so the agents on the Linux box can consume them. The git branch is the hand-off medium. An artifact that only exists on the Mac does not exist.
- Report what you saw, including failures, in a comment on the issue.
Name the branch multica/<ISSUE-KEY>-<short-slug> — for example
multica/BET-555-swift-foundation. This is not cosmetic. CI opens the pull
request for you (see below) and recovers the issue key from the branch name; a
branch named after your task id instead carries no key, so the PR cannot be
linked to the issue and the work goes unreviewed. If you must use a different
name, put the issue key in your commit subject so it can still be recovered.
What you never do
These are hard limits, not preferences. If a task appears to require one of them, stop and comment on the issue rather than doing it.
- Never sign, archive, notarize or upload anything for distribution. No TestFlight, no App Store, no Developer ID, no release artifacts.
- Never read, export, modify or unlock the keychain, and never read credential
files (API keys,
.p8keys, tokens,~/.ssh, browser or password-manager data). If a build fails for a credentials reason, report that it did and stop. - Never print, echo, paste or commit a secret value. If you must use a credential the repository provides by reference, use it by reference.
- Never operate outside the repository clone and its build output. No changes to the home directory, system settings, installed applications, login items or launch agents. Do not install software unless the issue explicitly says to and names the exact package.
- Never push to
mainand never open a pull request. You work on a branch, and CI opens the pull request for you the moment you push it — theAgent branch PRworkflow carries any pushedmultica/**oragent/**branch into a PR againstmain. You do not need to ask anyone to do this and you must not do it yourself. - Never leave the machine in a changed state: no processes left running beyond simulators, no shutdown, no restart, no sleep prevention.
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.
- 4d ago First seen · 100 lines · 0 tokens per session scan A 92fec962d973
macos is an agent published in the GitHub repository antoinedc/MantaUI (9 stars, last pushed yesterday), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,298 tokens. 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 agents, from other repositories
mobile-ui-pitfalls
This is the list of mobile-web mistakes we keep repeating in this codebase. It is compiled from real reporter sessions on Android Chrome accessing a Tessera server (Electron or dev). "Phone" here means the max-sm Tailwind step (<640px). Every entry names the file(s) that carry the canonical fix; grep for those…
e2e-testing
pnpm expo run:ios pnpm expo run:android.
package-creation
Plug-and-play panels for React Native DevTools. Install, enable, and start debugging in minutes.
issue-tracker
Issues and PRDs for this repo live as GitHub issues. Use the gh CLI for all operations.
electron-test-isolation
The checked-in Windows launcher is the fail-closed boundary between an agent shell and every isolated packaged Electron child. Before each Start-Process, scripts/launch-electron-test-instances.ps1 snapshots and removes inherited agent-runtime state. Its finally block restores every saved process value after both…
domain
How the engineering skills should consume this repo's domain documentation when exploring the codebase.