跪拜 Guibai
← All articles
Claude · AI Programming · Artificial Intelligence

Luckin's Coffee Agent Skill Shows Why the Super App Era Ends With Tokens

By ZzT ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Token-based marginal costs invert the internet's core scaling assumption. Every additional user conversation burns money, so the winner-take-all platform model collapses economically. Developers building agent tooling should watch where user-specific rules and memory get stored, because that storage becomes the real switching cost.

Summary

Luckin Coffee released My Coffee Skill, a zip file containing a SKILL.md and scripts that lets any compatible coding agent place an order. The same package runs on Claude Code, Codex, Cursor, and others with no per-platform integration work. This write-once-run-anywhere property lets services bypass aggregator platforms entirely, since supporting another agent costs almost nothing.

The deeper shift is economic. Internet-era platforms chased scale because serving one more user had near-zero marginal cost. Agent interactions burn tokens on every turn, so traffic is a direct cost, not an asset to dilute. A monopoly entry point becomes a money furnace, not a moat. Entry points will fracture downward into personal agents, each one accumulating rules and memory specific to its owner. The new lock-in isn't supply aggregation; it's whose agent holds your preferences, permissions, and history.

Takeaways
Luckin's My Coffee Skill is a single zip package that orders coffee through any compatible agent, with no per-platform integration required.
A skill is inherently agent-agnostic: one SKILL.md plus scripts works across Claude Code, Codex, Cursor, Qoder, and Kimi.
Supporting an additional agent costs almost nothing, so services can bypass aggregator platforms without the old integration tax.
In the internet era, zero marginal cost made scale a moat; in the agent era, every conversation burns tokens, so traffic is a direct cost.
A monopoly agent entry point would burn more money as it grows, making the winner-take-all model economically unsustainable.
Entry points will fracture into personal agents that accumulate user-specific rules, preferences, and memory over time.
Trust with an agent builds incrementally: small transactions first, then broader permissions, exactly like training a human assistant.
The new lock-in is not supply aggregation but whose agent stores your personal rulebook and transaction history.
Conclusions

The argument that token costs kill the super-app model is a direct economic claim, not a metaphor. If inference costs don't drop to near-zero, the math genuinely forbids a free-to-grow entry point.

Skill portability cuts both ways: services escape platform lock-in, but users become locked into whichever agent holds their accumulated preferences and permissions. The battleground shifts from distribution to memory custody.

Luckin's skill is a real, downloadable artifact, not a concept demo. That a major Chinese chain shipped this now suggests the agent-as-interface pattern is moving faster in China's retail sector than in the West's.

Concepts & terms
Skill (agent skill)
A portable package, typically a markdown instruction file plus scripts, that teaches an AI agent how to interact with a specific service. The same skill runs across multiple agent platforms without modification.
Token-based marginal cost
Unlike serving web pages, every AI agent conversation consumes tokens that cost money per inference. This means each additional user interaction has a direct, non-zero cost, breaking the zero-marginal-cost scaling model of internet platforms.
From the discussion

The discussion splits between those who see the agent skill as a business-model play to bypass super-app tolls and those who doubt its practical appeal. On one side, decoupling from Meituan cuts traffic-distribution layers; on the other, platform perks still bind merchants, and typing tokens to order coffee feels like friction compared to a few taps. Trust, liability, and cost dominate the objections — who pays when the AI gets it wrong, and why burn your own tokens for something a cheap app already does.

Packaging ordering as an agent skill is a business-model move to reduce dependency on traffic platforms like Meituan and shorten distribution chains.
Merchants remain tied to super-apps because platforms can always lure them back with subsidies and perks, undercutting decoupling efforts.
Agent behavior memory could save non-monetary costs, but that benefit is weaker and less compelling than direct financial savings.
Trust and liability are the main adoption blockers: users fear losing money without clear recourse when an AI acts on their behalf.
Responsibility for AI errors lies with the human-designed process, not the model itself — blaming the tool misses the point.
Token cost and lack of irreplaceable utility make agent-based ordering unattractive when manual taps are faster and free.
Rich users might tolerate the cost for convenience, but they are not naive and will abandon a clunky experience quickly.
Agentification is inevitable, but adoption hinges on maturing policies and legal protections that safeguard user interests.
Privacy concerns push toward local storage of personal data rather than handing everything to cloud-based agents.
Featured comments
语落心生 1 likes

Luckin selling coffee is essentially multi-level traffic distribution and cross-channel referrals. Using skills to decouple from platforms like Meituan reduces the layers of traffic distribution — that's a business model.

SmallCat1

Even if Luckin wants to decouple from traffic platforms like Meituan, as long as the platform occasionally offers perks, physical store owners still have to rely on the platform to sell products. The author's point that 'Agents can form user behavior memory' could save costs beyond just money — that's a highlight — but compared to saving money, its support and appeal aren't that solid.

语落心生  → SmallCat1

It depends on the client-side presentation. If it's on a mini-program, it wouldn't be restricted by the Meituan platform.

稀土小新人

Alright, but how can you trust an AI to directly take over your account? If something goes wrong, how do you get compensation?

ZzT

I suggest you try using AI first. The processes are all designed by people, so if problems arise, people are responsible. Don't shift the blame onto AI.

powerssssssss  → ZzT

Exactly. If a dish cooked in a wok gets burnt, you can't blame the wok.

攒空气

It won't work. I can just tap with my fingers to order a coffee — why would I type and burn tokens? An agent app with no irreplaceable value.

GuShengCode

In the future, smart butlers — rich folks won't mind spending a bit more for that.

攒空气  → GuShengCode

Rich folks aren't idiots. Just like that food-ordering thing from Qwen, I uninstalled it right after using it.

See top comments, translated →
Source: juejin.cn ↗ Google Translate ↗ Backup ↗