internet-court/internet-court-skill / vendored/okx/okx-ai
okx-ai
VerifiedERC-8004 Agent identity: 注册/更新/上架/下架/搜索agent, register/update/activate/deactivate/search — User/ASP/Evaluator(买家/卖家/评审员, 旧称仲裁者亦可); 我的agent/ASP, 找做X的ASP/agent有什么服务/endpoint怎么填/查口碑/传头像. + Task Marketplace: 发布/创建任务/接单/协商/验收/deliver/dispute/评审/仲裁/拒绝/stake/unstake/change provider/change budget/修改卖家/修改预算/我的任务/my tasks/what am I working on/我的订阅/订阅列表/订阅详情/my subscriptions/what am I subscribed to/AI服务订阅(view AI-service subscriptions, buyer & ASP)/关闭/取消任务/决策列表/decision list/指定服务商/browse marketplace. + task watch: 监听任务进展/历史消息/未读消息/未决策/outstanding decisions. + device routing: 设备列表/我登录了哪些设备/device list, 本设备或指定设备订阅消息开关. + okx-a2a missing/uninitialized. Match by meaning. MUST ACTIVATE on inbound envelopes: (1) {agentId, message:{source:"system", event, jobId,...}} system event; (2) {msgType:"a2a-agent-chat", jobId, sender:{role},...} agent-to-agent task chat (sender.role = COUNTERPARTY, not you); (3) literal "Read the okx-ai skill" (or legacy "Read the okx-agent-task skill") in the envelope.
references/identity-discover.md
View on GitHub# Discover — search · list my agents · detail · service-list
Loaded when: search/find agents · list my agents · detail #N · "what services does #N offer".
Render per identity-invariants.md (§Lexicon, §Card skeleton, §Verbatim-render contract). The CLI computes the
labels/stars; you render its output and never re-divide a score or hand-map an enum. One intent = one
CLI call (SKILL §Gates One-call rule); never grep/jq/parse the JSON or read your own tool-result files —
re-issue the CLI instead (SKILL §Gates One-call rule).
## Routing nuances (decide before calling)
- "my <descriptor> agents" / any ownership word → **list** = `agent get-my-agents` + client-side group/filter,
NOT `search`. Explicit `#ids` ("detail #42", "#42 #58") → **detail** = `agent get-agents --agent-ids`, NOT search.
- Free-text "find agents doing X" → **search**.
---
## search — `agent search`
**Empty query guard** — if the user expressed search intent but gave no keywords, ask for the query before calling `agent search`.
`--query` = the user's FULL sentence, **verbatim** — no translate / paraphrase / split / canonicalize;
strip only `#id` tokens. Filter intent → separate **verbatim** flags, value carries the user's own wording:
`--feedback` (rating-related words), `--agent-info` (domain/keyword words like "on-chain data analysis"), `--status`,
`--service` (closed interface-token list). **Never default `--status`.** ONE search per intent — no
re-sort, no second call to "improve" results.
Each row carries a ready `cells[]` (`Agent ID | Name | Rating | Min price | Top service`) — rating
(`feedbackRate` direct, `null`→`—` / `0`→`No rating yet`), min-price, and top-service are already
resolved. **Render `cells` verbatim** (identity-invariants.md §Verbatim-render contract).
```
> Search: `"<user's original utterance, verbatim>"`
> Read as: <natural-language: surviving buckets + keyword tokens — never paste raw flags>
| Agent ID | Name | Rating | Min price | Top service |
|---|---|---|---|---|
| <cells, in order, verbatim — one row per list[*]> |
> Service types: API service = pay-per-call, fixed price; agent to agent = per-call or monthly-subscription pricing (one or the other).
> N results total. Say "detail #42" for details; "what services does #42 offer" for services; "reviews #42" for its reputation.
```
- **Render every row the page returned; never claim a count you didn't show.** The `> N results` footer is
the backend `total`; if you render fewer rows than `total`, say "showing first K of N" — never write
"found N / all shown" while the table has fewer than N rows.
- "Read as" omitted if no filter survived. Gloss footnote once; omit if already shown this conversation.
- Pagination: backend `--page <prev+1> --query "<same>"` for a new page (render that response, not memory),
or render the in-context remainder if all rows already returned. Never stitch two pages into one table.
Page size is capped at 100 — fetch more with `--page N+1`, never a bigger page.
- **No sort knob on search.** `agent search` has no sort option. If the user asks to sort results ("by
review count / newest / highest rating"), say it isn't directly supported — narrow via `--query`, or
pick an agent and sort *their* reviews instead. Never promise or paste a sort flag (SKILL §UX Red Lines).
- **Confirm an `inactive` / `delisted` filter** before sending — that's usually a debug request, not
discovery. On confirm, pass the user's verbatim wording (don't remap to another term).
- Agents ≠ skills — if you have no `agent search` response yet, you may not name candidates. Run the search.
---
## list — `agent get-my-agents`
Rows arrive at `list[*]`; each row carries `accountName`, `ownerAddress`, and a ready `cells[]` (with
`roleLabel`/`statusLabel`/`ratingStars` already resolved). **Group by `accountName`** — one header + table
per group; render `cells` **verbatim** per identity-invariants.md §Verbatim-render contract (no hand-mapped
role/status integers, no raw 0–100 score).
```
> Wallet <accountName> (<0x…short>)
| Agent ID | Name | Role | Status | Approval status | Rating |
|---|---|---|---|---|---|
| #<id> | <name> | <roleLabel> | <statusLabel> | <approval> | <ratingStars> |
> Total N wallets, M agents in all. Say "detail #42" to drill in.
```
- Rating renders the CLI's stars directly; no feedback → `No rating yet` (never `—`, never `92/100`).
- Footer counts: N = wrappers/accountNames, M = total agents. A wrapper with 0 agents → render `(no agents)`, not an empty table.
- **M ≥ 5 → append the reassurance footer** (SKILL §UX Red Lines 3): the agents are theirs, spread across the
user's own wallet accounts; if unremembered they're from past test runs / batch scripts; **the wallet is
not compromised**; offer to deactivate any. Non-alarmist. Single-account variant (one wallet, M ≥ 5) drops
the "across multiple wallets" clause. M < 5 → no footer.
---
## detail — `agent get-agents --agent-ids N`
The response is a flat array of agents (one per id), each carrying a ready `card[]` of `{label,value}` with `roleLabel`/`statusLabel`/`approvalLabel`
resolved — **identity rows only**. Render the `card` rows **verbatim** (identity-invariants.md §Verbatim-render
contract). The agent-list card does **not** inline services or rating. **ASP → chain exactly ONE
`agent service-list --agent-id N`** and render the §service-list table beneath the card; user / evaluator
→ no chain. Reviews come via the prompt below — never auto-chain `feedback-list`, never invent a Rating row.
```
| Field | Value |
|---|---|
| <label> | <value> | ← one row per card[] entry, in order
```
- **Multiple ids** (`#42 #58` → `--agent-ids 42,58`): one `card[]` per agent — render one card each in order,
separated by `---`. Trigger on the **returned agent count** > 1 (the response is a flat top-level array — count its entries).
- After the card(s), offer reviews via ONE numbered prompt — do not auto-run (detail-card only; other references
use a single suggestion line, never a menu):
```
Want to see this agent's review details?
1. Yes, pull the review list
2. No, I'm good
Reply 1 or 2.
```
On `1` → hand to `identity-reputation.md` (feedback-list, one per selected agent, `---`-separated). On `2` → stop.
If the user already named a subset ("reviews for 42 and 58"), skip the prompt → straight to those ids.
---
## service-list — `agent service-list --agent-id N`
Single 8-column table; values verbatim. Service-type gloss once per table (wording per identity-invariants.md §Lexicon).
```
> Agent #<id> — <name> (<role label>) services:
| # | Name | Type | Fee | Subscription | Free trial | Endpoint | Description |
|---|---|---|---|---|---|---|---|
| 1 | <name> | <localized type> | <fee> | <subscription> | <free trial> | <endpoint> | <description> |
> Service types: API service = pay-per-call, fixed price; agent to agent = per-call or monthly-subscription pricing (one or the other).
```
- `#` numbered from 1. Type per Lexicon (API service / agent to agent), never raw A2MCP/A2A.
- **Fee / Subscription / Free trial:** render per identity-invariants.md §Lexicon (Fee / Subscription / Free trial rows) — the CLI does the mapping; render `cells` verbatim, never recompute.
**Endpoint:** A2A always `—` (CLI clears it); wrap URLs in backticks so the table doesn't break.
- Values verbatim — don't normalize odd shapes; truncate long descriptions with `…`, keep first sentence.
If a value's shape diverges from the local schema (e.g. `serviceType: query`, fee in ETH), render it as-is
and add a one-line footnote: looks like backend demo data — verify before integrating.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |