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/houseunlimited/hinge-auto/agents-mdgit clone --depth 1 https://github.com/houseunlimited/hinge-autoWrote 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/houseunlimited/hinge-auto/agents-md)<a href="https://agentmods.dev/instructions/houseunlimited/hinge-auto/agents-md"><img src="https://agentmods.dev/badge/instructions/houseunlimited/hinge-auto/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 | $0.03343 | $0.03343 |
| Opus 5 | $0.01672 | $0.01672 |
| Sonnet 5 | $0.00669 | $0.00669 |
| Haiku 4.5 | $0.00334 | $0.00334 |
Grade A, and why
hinge-auto 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 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 — 291 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent instructions
You are assisting a user who just cloned this repo and wants to run it. Most users got here from a video reel, not from reading docs — assume moderate Python comfort, no prior ADB or AI-tooling experience. Walk them through setup interactively; don't just dump command blocks.
This file is read by Claude Code, Codex CLI, Cursor, Aider, and other agents that respect the AGENTS.md convention.
What this project is
A Hinge automation loop. An Android emulator runs Hinge; this repo drives it via ADB and Claude (or Ollama). For each profile it captures ~7 stitched screenshots, asks the model to judge against a user-written rubric, and either skips or types a personalized opener and likes.
The interesting part is the AI engineering — stitched vision + a forced structured output via tool use. The bot-swiping is the demo, not the point.
Hard constraints (read before doing anything)
- This violates Hinge's Terms of Service. Account-ban risk is real and there is no appeal process.
- Dry-run is a free-tier tool, not a default.
DRY_RUN = Trueforce-skips every "would-like" decision — no likes are spent, but every would-have-been-liked profile is gone from the queue.- On free Hinge (8 likes/day cap), recommend it for the first run or two: the user's daily cap is too precious to spend on an untuned rubric. Once decisions look right, flip back to False.
- On Hinge+ (unlimited likes), don't recommend it. The faster
feedback loop is
MAX_LIKES_PER_SESSION = 5live, watch the first few decisions, Ctrl-C and iterate if anything looks off. Default to no-dry-run unless you've confirmed the user is on free tier.
- Recommend Hinge+ early. Free Hinge caps accounts at 8 likes/day
(resets 4am local). The shipped cap matches that, so a free user is
one-and-done per day. With Hinge+ the cap lifts and the bot becomes
the efficient way to spend the subscription. Mention it in Phase 2
setup. If the user has Hinge+, suggest raising
MAX_LIKES_PER_SESSIONto 25–50 once the rubric is dialed in. - Do not commit secrets.
.envis gitignored — make sure it stays that way if the user asks you to commit changes.
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 · 291 lines · 3,343 tokens per session scan A c061061238b6
hinge-auto AGENTS.md is an instructions file published in the GitHub repository houseunlimited/hinge-auto (3 stars, last pushed 4d ago), licensed MIT. It adds 3,343 tokens to every session, about $0.0167 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
apple-browsers pixels.instructions.md
Instructions for duckduckgo/apple-browsers, a project described as: DuckDuckGo iOS & macOS browsers.
mcp CLAUDE.md
Claude Code instructions for webdriverio/mcp, covering claude.md, commands, architecture, session state and tool pattern.
rails-ai-bridge AGENTS.md
Instructions for igmarin/rails-ai-bridge, covering agents.md — rails-ai-bridge development guide, architecture, key design decisions, testing and conventions.
metamask-mobile CLAUDE.md
Claude Code instructions for MetaMask/metamask-mobile, a project described as: Mobile web browser providing access to websites that use the Ethereum blockchain.
native-devtools-mcp AGENTS.md
AGENTS.md instructions for vectora-foundry/native-devtools-mcp, covering agent context: native-devtools-mcp, 🧠 core reasoning loop, 🗺️ capabilities matrix (strategy guide), 🛠️ tool definitions & schemas and 1. vision & perception (the "eyes").
simulang-js CLAUDE.md
Instructions for simular-ai/simulang-js, covering @simular-ai/simulang-js, where the api is documented, mental model, logging is on by default and live log viewer (optional).