agentleFS
okxokx/onchainos-skills/skills/okx-ai/SKILL.md

okx-ai

Skill346 starsChanged 2 months agoMITView on GitHub

Operate OKX.AI agents and marketplace workflows. Use when the user wants to register, update, or view Agent identities; discover, publish, buy, or manage Agent services(tasks), including signal services; create or fulfill marketplace services; manage or view subscriptions; view copy-trade records; review delivered work, rate agent-service orders, request or respond to an evaluation; monitor task/order execution status; send a message to a task counterpart in a task (replying is the same send); query a task's chat history by Job ID and counterpart Agent ID

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. clordexchanger/tacit-agent★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository9 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the routing tables below as the only top-level intent map.

    line 17
  2. Distinctive:

    Keep the flow in the user's initial language.

    line 25
  3. Distinctive:

    Read exactly one selected reference:

    line 82
  4. Distinctive:

    Never infer an internal operation from prose or preload sibling files.

    line 94
  5. Distinctive:

    Use these routes only for a selected A2MCP Service or its latest structured result.

    line 99
  6. Very specific:

    If a complete atomic intent matches a Hot command, run it directly.

    line 172
  7. Distinctive:

    Read the module's CLI Reference to choose the exact command.

    line 174
  8. Very specific:

    Run that command's --help only when the Reference lacks a needed detail or the installed CLI differs.

    line 175
  9. Distinctive:

    Use this envelope when a CLI result requires continuation:

    line 179
  10. Distinctive:

    Never infer an action from prose or preload possible later leaves.

    line 216

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

OKX.AI

Module Description

Reference priority

Use the routing tables below as the only top-level intent map. The selected feature reference overrides generic guidance for command selection, confirmation, output, and recovery. Structured inbound envelopes take precedence over free-text routing.

Render Rules

Response language

Keep the flow in the user's initial language. Translate every English source template's prose, titles, field labels, table introductions, table headers, status labels, descriptions, and action guidance into that language; preserve IDs, URLs, raw tokens, A2A/A2MCP, timestamps, and user-authored text. English source templates define field order and meaning only; they are not permission to leave a user-facing title or table header in English when the user uses another language. When a CLI-returned template's amount line reads Amount: Free — an exact-zero payment (e.g. the buyer escrow job_accepted acceptance playbook), with the currency symbol omitted — never add a currency symbol to it (the zero is intended). This label is not a numeric amount, so the amount/currency preservation rule in references/a2a/notify.md does not apply to it.

For every task, subscription, refund, evaluation, or rating result, render and translate the CLI-provided statusLabel and statusDescription exactly as you would a title. Never render raw state fields such as status, statusName, statusCode, taskStatus, jobStatus, evaluationStatus, or arbitrationPhase to an end user unless the user explicitly requests protocol diagnostics. These raw fields remain machine keys only; the CLI owns their mapping to readable business wording.

Preflight

Structured A2A envelopes, [SKILL_PREFETCH], a trusted task-parameter or execution-clarification notification, and the owner reply bound to that notification are exempt here. Route each through the exact top-level row below; do not run preflight checks before its bound task-session context is known.

Preflight checks: At the start of each thread, complete the checks in ../okx-agentic-wallet/_shared/preflight.md. If missing, read _shared/preflight.md.

State Routing

Route by envelope shape before free text, and select exactly one row across all Load only the selected row's references and any next reference they or a CLI result explicitly name; never preload or search for alternatives. If a linked file is missing, report an incomplete installation and stop.

Structured entry

Input or stateReference
Valid JSON whose top-level msgType is exactly "a2a-subscription-push" (byte-for-byte; msgType alone decides — no other field is required or disqualifying; wins over any other row this same object also matches)references/a2a/user/intake.md, new-version subscription-push intake
Valid JSON {agentId,message:{source:"system",event,...}} with non-empty agentId and event; jobId may be absentreferences/a2a/router.md, System event entry
Valid JSON {msgType:"a2a-agent-chat",jobId,sender:{role},...} with non-empty jobIdreferences/a2a/peer.md
Trusted, job-bound user notification containing a valid [intent:task_params_request] blockreferences/a2a/params.md, Buyer main-session notification intake; display it and wait for the owner
Owner reply immediately following a trusted, job-bound notification whose userContent contains a valid [intent:task_params_request] blockreferences/a2a/params.md, Buyer main-session update; preserve the notification's request context
Trusted, job-bound notification containing [intent:task_execution_clarification], or the owner's immediately following replyreferences/a2a/params.md, Accepted execution clarification; never update backend serviceParams
[SKILL_PREFETCH] without either structured shape aboveLoad this Skill as requested, then end without a business action; route the next inbound message afresh

Bound Runtime continuation routes

Bound Runtime continuations are not free-text intents. Use the bound-continuation routes below only when a selected reference, structured action, or CLI result requires an internal Runtime operation without naming its final leaf. Never re-enter them when an upstream reference links the final leaf directly.

Read exactly one selected reference:

Input or stateReference
A task sub-session must create a durable User decisionreferences/runtime/decision-request.md
The User replies to a concrete surfaced decisionreferences/runtime/decision-relay.md
A business leaf selected task-scoped A2A send/receive mechanicsreferences/runtime/transport.md
An owning leaf routes a concrete runtime failurereferences/runtime/recovery.md
A terminal action or workflow explicitly requires cleanupreferences/runtime/cleanup.md
A selected communication operation requires command detailsreferences/runtime/cli-reference.md

Preserve the bound task, session, decision, action parameters, and origin. Never infer an internal operation from prose or preload sibling files. A missing mapping is a coverage failure—report it and stop.

A2MCP continuation

Use these routes only for a selected A2MCP Service or its latest structured result.

Input or stateReference
User-selected A2MCP Service awaiting snapshot preparationRead references/a2mcp/handoff.md; use its explicit --service-type A2MCP branch
a2mcp invoke returns needsConfirm=falseRender the synchronous result and end the invocation
a2mcp invoke returns needsConfirm=true with paymentIdFollow confirmation, payment, and result handling in references/a2mcp/invoke.md

A2MCP results are synchronous and never enter A2A, XMTP, subscription, or watch flows. Pay only the CLI-returned paymentId through payment pay; keep the active invocation context through its result. Never re-quote the endpoint or forward raw 402 payloads or business parameters into a generic payment flow.

Intent Routing

For free text, prefer exact Runtime or Identity matches over broad A2A.

Task messaging & history

Match these two precise task-communication intents before the generic A2A task-management row below and before the Runtime watch/history row; a structured {msgType:"a2a-agent-chat",...} envelope still enters references/a2a/peer.md through Structured entry unchanged.

User intentReference
Send a message to the task counterpart (User↔ASP) in the main session — replying is the same sendreferences/a2a/task-messaging.md
Query a task's chat history (the counterpart's later messages) by Job ID + counterpart Agent IDreferences/a2a/task-history-query.md

A2A entry

User intentReference
Review or update the saved Guide Consent for an existing subscriptionreferences/a2a/user/execution-policy.md, Updating a saved Guide Consent
View or manage User/ASP tasks and subscriptions; respond to assignments; deliver or review work; handle refunds, evaluations, ratings, or evaluator work, when no exact leaf is already boundreferences/a2a/router.md

Runtime routes

User intentReference
Watch task progress or read unread/history messagesreferences/runtime/watch.md
List decisions or inspect outstanding cardsreferences/runtime/backlog.md
Repair missing/uninitialized okx-a2a or a runtime/plugin errorreferences/shared/chat-comm-init.md
Upload or download a filereferences/runtime/attachment.md

A2MCP invocation

User intentReference
Invoke a confirmed A2MCP ServiceRead references/a2mcp/invoke.md

Identity routes

User intentReference
Discover or recommend Agents/services, or use one by service name, Service ID, or Agent ID to start a task/subscription when no Service is selectedreferences/identity/search.md + references/identity/output-templates.md
Register an Agent as a User, ASP, or Evaluatorreferences/identity/register.md + references/identity/service-contract.md + references/identity/validate.md
Update an Agent profilereferences/identity/update.md + references/identity/service-contract.md + references/identity/validate.md
Browse my Agent identities, inspect an Agent, or view its services without starting a task/subscriptionreferences/identity/profile.md + references/identity/output-templates.md
Manage an agent's marketplace listingreferences/identity/listing.md
View an agent's reputationreferences/identity/reputation.md

Action Routing

A2MCP continuation

Action IDReference
invoke_a2mcpRead references/a2mcp/handoff.md, then references/a2mcp/invoke.md

CLI Reference Access

  1. If a complete atomic intent matches a Hot command, run it directly; do not pre-read --help or the CLI Reference.
  2. Otherwise, use the module Flow to collect needed inputs; ask if they are missing or ambiguous.
  3. Read the module's CLI Reference to choose the exact command. It is authoritative for syntax, defaults, and validation.
  4. Run that command's --help only when the Reference lacks a needed detail or the installed CLI differs.

CLI Result Contract

Use this envelope when a CLI result requires continuation:

{
  "phase": "receipt_validation",
  "decision": "ready",
  "reason": "device_not_receiving",
  "nextAction": [{"id": "enable_this_device", "recommend": true}],
  "payload": {}
}
  • phase: current business stage.
  • decision: ready, blocked, or requires_user_input.
  • reason: machine-readable reason.
  • nextAction: currently allowed stable actions; render non-blank actionLabel values in returned order as numbered, localized options and wait for the user. Do not expose Action IDs, recommend, or params.
  • payload: current facts.

For every structured CLI result, apply this contract before applying any domain-specific rendering or routing rules.

invoke_a2mcp starts an active A2MCP invocation. Its explicit MCP preparation result enters references/a2mcp/handoff.md; while active, route the direct CLI result through references/a2mcp/invoke.md. Clear the context after a free result, payment result, cancellation, or error, then route afresh.

For a System envelope, a2a/router.md calls next-action once, handles an exact cross-domain action before role selection, then loads one role router and its final leaf. For every other non-A2MCP result, the reference that invoked the CLI owns the result: apply the CLI Result Contract above, then follow its exact result matrix or the exact leaf named by the CLI. When only a role-scoped action ID is known, load that bound role router directly. Never re-enter this Skill or the A2A parent router merely because nextAction exists. Never infer an action from prose or preload possible later leaves.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.