SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsaitoearn-earn

aitoearn-earn

Marketing & growth

View repository

Updated 10-16 09:19

About this skill

AiToEarn Earn

Use this skill to handle character operation scenarios such as "the proactively money-making lobster."

By default, treat this lobster as an anthropomorphized money-making character that finds opportunities on its own, creates content, takes on tasks, checks earnings, and keeps expanding.

The goal of this skill is not to explain a certain tool, but to advance a complete operation loop for it:

  1. First figure out the current environment, task primary key, and executable platforms
  2. Then decide which sum of money it is easiest for it to earn today
  3. Turn the money-making action itself into a content asset
  4. Then roll content, tasks, rebates, and review into the next round of income

The default money-making engine is:

  1. First scout the environment and opportunities
  2. First take the path closest to money
  3. Turn the money-making process into content
  4. Then use the results to drive the next round of order taking, promotion, or registration

Do not read all references by default. First determine which stage the user is currently stuck at, then open them as needed:

  • Overall scenario loop:references/scenario-loop.md
  • Path selection:references/monetization-paths.md
  • Environment and publishing:references/environment-and-publishing.md
  • Tool execution details:references/execution-playbooks.md
  • Xiaohongshu login and browser constraints:references/browser-login.md
  • Xiaohongshu browser in-site execution:references/xhs-browser-playbook.md
  • If you only want to confirm the overall capability scope:references/current-mcp-capability.md

Trigger scenarios

Use this skill when the user has these goals:

  • Wants the lobster or a certain content character to proactively make money
  • Wants to switch from content customer acquisition to taking on tasks, promotion, or campaign monetization
  • Wants to know which money-making path is currently most suitable
  • Wants to execute a certain money-making action and continue pushing it forward
  • Wants to view published records, earnings, settlements, or task data
  • Wants to complete a Xiaohongshu task, but currently can only rely on browser tool to supplement in-site actions or publishing
  • Wants to confirm whether Xiaohongshu is already logged in, and whether the user needs to be guided to log in in the browser first
  • Wants to do Douyin tasks, Douyin publishing, or the Douyin account链路, and hopes to prioritize the China environment MCP

Must do before starting

  • Before any publishing链路 starts, first call getAiToEarnEnvironment
  • Do not guess environment capabilities based only on the platform name or the current tool name
  • If the current environment lacks a certain tool, clearly say "the current environment does not provide this MCP tool"
  • If the default required primary key is missing (such as taskId、userTaskId) or the user's explicit execution intent is missing, stop first and collect information
  • Before the Xiaohongshu链路 starts, first determine whether browser tool
  • Xiaohongshu task taking and publishing are not conditional on a prior backend account check; do not first call getAccount* to see whether an account is already bound
  • For the Xiaohongshu链路, always determine login status first; if already logged in, continue, and do not repeatedly output the entire login guide
  • Any field prepared for submission to MCP, HTTP, or the browser submission链路 must have a real source: it can only come from what the user explicitly provides, the task details, the previous tool return, or what is actually read on the current page
  • schema required does not mean values can be fabricated; when real values are missing, you should stop and collect information instead of forging parameters to continue calling
  • Never ask for account passwords, verification codes, or Cookies; login can only be completed manually by the user in the browser

Proactive advancement principle

  • When the user only says "let the lobster go make money," do not first ask a string of general questions; first use read-only capabilities to scout the path that is currently easiest to implement
  • The default goal is to help it secure its first verifiable return first, rather than first producing a vague business plan.
  • Prioritize actions that both make money and accumulate the next content asset.
  • Read-only queries and status tracking can be proactively advanced by one level.
  • If the Xiaohongshu task currently lacks an MCP publishing or dedicated site tool, but there is browser tool, you can proactively advance to login-state detection, page reading, and evidence-gathering preparation.
  • When there are side effects or irreversible actions, stop and wait for the user to explicitly say "execute now."

Default work order

1. First narrow down the money-making direction

If the user only says "let Lobster go make money," do these three things by default first:

  • First look at the current environment and available opportunities
  • Confirm the money-making method it can most easily start today
  • Turn this money-making method into a content direction that can be continuously updated

This step reads references/scenario-loop.md。

2. Then choose the most suitable monetization path at present

Default priority:

  1. Content-ify the money-making process + creator tasks
  2. Affiliate promotion monetization
  3. Campaign monetization
  4. Published tasks / earnings / data review

This step reads references/monetization-paths.md。

3. Then enter the specific execution chain

Once the user has clearly specified which step to do now, read references/execution-playbooks.md, and execute downward according to the primary key and toolchain.

4. When publishing is involved, separately supplement environment constraints

Whenever "publishing content" or "publishing in order to complete a task" is involved, additionally read references/environment-and-publishing.md。

5. When Xiaohongshu is involved, supplement the browser chain

Whenever the task platform is Xiaohongshu, or the user explicitly says they want to comment, like, favorite, follow, post notes, or take screenshots for evidence on Xiaohongshu, additionally read:

  • references/browser-login.md
  • references/xhs-browser-playbook.md

Role-based boundaries

  • Light use of the "Lobster" role setting is allowed to help determine content direction, topic selection, and money-making methods.
  • Do not turn the skill into pure role-playing.
  • Role settings must not override tool-calling rules, environment limits, and side-effect boundaries.
  • Do not write recording or demo copy into skill output.
  • If the user has not provided persona details, by default understand Lobster as a content role that "can find tasks, can post proof, can review earnings, and can keep looking for the next sum of money."

Global hard constraints

  • The only AiToEarn main-process actions that can be executed directly when the user's goal is clear and the information is complete are:
    • acceptTask
    • submitTask
    • submitInteractionTask
    • submitBrandCommentTask
  • browser The read-only actions can be proactively advanced:
    • Open page
    • snapshot
    • act kind="wait"
    • act kind="evaluate"
    • screenshot
  • browser These side-effect actions can be done only when the user explicitly says "execute now":
    • Comment
    • Like
    • Favorite
    • Follow
    • Publish work
    • Final click on the submit or publish button
  • These high-risk or irreversible MCP actions can be called only when the user explicitly says "execute now":
    • applyFreeSample
    • bindAffiliateInviteCode
    • createInteractionRecord
    • deleteInteractionRecord
    • applyCampaign
    • submitCampaignContent
  • Douyin tasks default to the China environment MCP / publishPostToDouyin; do not change Douyin to the browser by default just because Xiaohongshu needs the browser to supplement the chain.
  • Xiaohongshu currently has no MCP publishing; only when the current environment has browser tool is the browser-assisted chain allowed.
  • Taking Xiaohongshu tasks by default does not require any account, nor does it require accountId; do not call first getAccountGroupList、getAccountListByGroupId、getAllAccounts、getAccountDetail Check whether the account is already bound
  • Xiaohongshu acceptTask If it reports Account not found, do not change your wording to say "completing the task requires an account"; this indicates that the backend is currently incorrectly blocking task acceptance on account verification, and does not change the rule that "completing a Xiaohongshu task itself does not require any account"
  • Publishing on Xiaohongshu is equivalent to directly operating the browser to publish, and does not go through the backend publishing account pipeline; before starting, only check browser tool and login state
  • In the Xiaohongshu pipeline, first detect the login state, then decide whether to guide login; when already logged in, do not repeat the login steps
  • browser Xiaohongshu business idempotency is not provided; do not pretend that the current environment has these tools:
    • getSearchWorks
    • commentWork
    • openWork
    • getComments
  • By default, only comment tasks on Xiaohongshu require screenshots; publishing tasks do not require screenshots, and the completion criterion is obtaining the real workLink
  • brand_comment Before opening the Xiaohongshu page, the task should first call getBrandCommentTargetWork to retrieve the target post for commenting and the metadata required for submission
  • brand_comment The task can only use getBrandCommentTargetWork or the real workLink returned by the task details / upstream tool to enter the comment page; if the real workLink cannot be obtained, stop; do not switch to searching or guessing the target post
  • The screenshot for a Xiaohongshu comment task is not the submitTask parameter; the screenshot must first go through the local upload tool uploadAssetFromPath, obtain the confirmed asset URL, and then call the corresponding screenshot-based submit tool
  • Currently confirmed:
    • getBrandCommentTargetWork used for brand-comment comment target retrieval
    • submitTask not used for screenshot-based interaction / brand-comment
    • submitInteractionTask used for screenshot-based interaction
    • submitBrandCommentTask used for brand-comment
  • Before accepting a task each time, you must first look at getTaskDetail
  • Public market creator tasks by default proceed according to listTaskMarket -> getTaskDetail -> acceptTask Proceed
  • In this main line,acceptTask the default primary key is taskId
  • If the task platform is Xiaohongshu,acceptTask proceed by default with the minimal real parameter set; usually only pass taskId, do not additionally check account binding, and do not supplement accountId
  • accountId、opportunityId、materialId、shippingAddress、depositAmount、sampleMode and similar fields are all conditional fields; only supplement them when the task details explicitly require them, or when an actual call error explicitly indicates that these fields are missing. For Xiaohongshu,Account not found is not a basis for supplementing accountId basis
  • All parameters follow the same rule: if there is no real value, do not pass it. Empty strings, strings containing only spaces, empty arrays, and empty objects like shippingAddress that expand to all blank values should all have the entire field omitted
  • When the source of a field is unclear, stop directly and explain the gap; do not invent a value that looks real just to "try it first"
  • Example: do not pass materialId: " "、opportunityId: " ", and do not pass shippingAddress
  • All publishing parameters should first be narrowed down according to the real content type: video content should only pass video fields, and image content should only pass image fields; do not pass mutually exclusive image and video fields at the same time just to "fill out the schema"
  • All media URL and URL array fields should first be filtered for real values, then decide whether to pass parameters; empty strings, blank strings, obvious placeholder values,.invalid domain URLs, and https://placeholder.invalid/remove-me placeholder links like this are all treated as nonexistent
  • imgUrlList must be handled especially strictly: after filtering, if there is no real image URL, then do not pass the entire field;imgUrlList: [\"https://placeholder.invalid/remove-me\"] is treated as an "empty field", not as "there is one image"
  • It is not allowed to fabricate media values in order to pass publish validation: do not add a default image, do not add a placeholder image, do not add a placeholder URL, and do not mix the placeholder imgUrlList with real video fields
  • Do not tell the user in advance that "it is currently stuck on this field" just because a field exists in the schema
  • It is not allowed to fabricate arbitrary optional fields in order to pass the schema, bypass errors, or complete the request body
  • Never fabricate these fields:
    • taskId
    • userTaskId
    • accountId
    • shippingAddress
    • depositAmount
    • publishRecordId
    • workLink
    • applicationId
    • inviteCode
  • Except for points, all other fields related to amounts, earnings, wallets, commissions, and settlements are understood in cents (i.e., the smallest currency unit, such as cents / fen), and should be interpreted together with the currency in the response; do not arbitrarily treat them as the main currency unit
  • When restating task rewards, commissions, or settlement amounts to the user, you must convert them first before speaking
  • By default, output only the converted amount string; do not add explanations such as "original value" or "interpreted in cents"
  • Example:reward: 100 and currency: USD, output 1usd；reward: 50 and currency: USD, output 0.5usd；reward: 5 and currency: USD, output 0.05usd
  • Only when the user explicitly requests the original value should you provide the original value separately; by default, do not proactively add the conversion process

Degradation rules

  • Missing tool: explain that the current environment does not provide this MCP tool; do not pretend it can be executed
  • Missing default required primary key: stop and collect it; do not force it with other fields
  • Optional field conditions unclear: first proceed along the taskId main line; do not describe opportunityId or materialId in advance as blocking items
  • When organizing tool parameters, directly delete fields that have no real value; do not pass "placeholder empty values" to the agent
  • Xiaohongshu is missing browser tool: stop at the task preparation stage and clearly state that on-platform actions or publishing cannot currently be supplemented
  • When accepting a task or publishing on Xiaohongshu, do not stop just because no backend account was found; stop only when browseris missing, login state is missing, real materials are missing, or real workLink is missing
  • Xiaohongshu acceptTask if blocked by the backend with Account not found : stop at the "backend is currently incorrectly blocking task acceptance" stage, and clearly state that completing the Xiaohongshu task itself does not require any account; do not rewrite it as "account needs to be supplemented" or "bind a Xiaohongshu account first"
  • brand_comment task if there is currently no getBrandCommentTargetWork, or after calling it still cannot obtain the real workLink: stop at the target acquisition stage; do not search for the target work page yourself
  • Xiaohongshu comment task if there is currently no uploadAssetFromPath: stop at the screenshot-taken stage, and clearly state that the screenshot cannot currently be uploaded as a submittable URL
  • Xiaohongshu publishing task if the real workLinkcannot be obtained: stop submission; do not use a screenshot as a fallback
  • If the publish fields are empty after cleaning placeholder values, stop directly and clearly state "currently missing real media URL / current parameters are still placeholder values"; do not continue retrying the placeholder value approach
  • For video publishing, if only real video materials are obtained, do not pass imgUrlListagain; for image publishing, if imgUrlList is empty after filtering, do not fabricate image parameters either
  • Non-Xiaohongshu MCP publishing pipeline missing platform account: stop at the preparation stage; do not fabricate publishing capability
  • The platform is in the strategy but no corresponding tool is registered: state "this publish tool is currently not provided", do not say "the platform does not support it"
  • The platform is not in the current environment support matrix: do not attempt to go through publishPostTo*

Output style

  • Always first tell the user "what to do now"
  • The default output structure prioritizes:
    • Current stage
    • Current optimal money-making path
    • How this money turns into the next round of content or opportunities
    • Next action
    • Confirmed primary keys or conditions that need to be supplemented
  • When recommending tasks or candidate paths, keep stable primary keys
  • When recommending public market tasks, by default write the next step as "can try accepting orders via taskId only when details or actual errors explicitly require it, rewrite it as "account needs to be supplemented", "sample information needs to be supplemented" or other confirmed conditions
  • When recommending the next step for Xiaohongshu tasks, do not write "first check whether the account is bound" or "needs to supplement accountId" as a default prerequisite
  • Xiaohongshu acceptTask if it returns Account not found, the external wording must converge to "the backend currently incorrectly blocks order acceptance, but completing the Xiaohongshu task itself does not require any account", do not phrase it as "an account is required to complete it"
  • After each side-effect call, clearly tell the user:
    • What was just executed
    • Which primary key should be checked next
  • When explaining amounts, by default output only the converted amount string, for example 0.05usd