CRITICAL: MUST use the Read tool to read

../openydt-shared/SKILL.md

before starting to master openydt's configuration / profile / signature (v2/v3) / response envelope and status codes / rate limiting / write operation security rules. All newly created Skills must reuse this base; do not repeat this content in new Skills.

lark-skill-maker

SKILL.md

• Users want to create a new openydt domain Skill (e.g., "on-site vehicle / monthly pass / blacklist" domains).
• Users want to encapsulate a specific

  openydt api <cmd>

  or a business link like "check fee → pay fee" into a reusable Skill.
• Users want to standardize an existing SKILL.md (supplement frontmatter, organize command tables, add

  --yes

  to write operations).

openydt api

cmd/gen

---
name: openydt-<domain>                 # Required, same as directory name, kebab-case
version: 1.0.0                      # Required, semantic versioning
description: "<one-sentence responsibility>. <enriched trigger words>."  # Required, see requirements below
metadata:
  requires:
    bins: ["openydt"]              # Required, declare dependency on openydt binary
  cliHelp: "openydt <domain> --help"   # Recommended: allow users to quickly view real subcommands
---

• description

  is the key to correct recall of the Skill, must be rich in Chinese trigger words: first use one sentence to clarify the responsibility (which domain it covers, whether it is read or write), then list possible colloquial expressions of users (e.g., "check parking lot, empty parking space, check fee, issue coupon, add to blacklist, activate monthly pass" and other synonyms/near-synonyms), spread out read/write verbs and business nouns. Refer to the description density of

  openydt-park

  /

  openydt-coupon

  .
• Fill in the real help command of the domain for

  cliHelp

  to facilitate humans to verify whether subcommands exist.

1. CRITICAL: Read shared first: The first sentence of the main content must be a prominent reminder requiring users to Read

   ../openydt-shared/SKILL.md

   before starting, and not repeat signature/status code/security rules in this Skill.
2. When to use this Skill + Intent Routing: Clarify what this domain is responsible for, and provide routing like "please switch to X domain for such requests" (to avoid Skill overstepping).
3. Available Command Table: List the domain commands in a table with columns "Chinese Name | Command | Read/Write | Key Parameters".
   • The

     Command

     column writes real executable

     openydt <domain> <use>

     .
   • The

     Read/Write

     column indicates read or write; all write commands must be marked with "requires

     --yes

     " in the table and examples.
   • Mark required items in

     Key Parameters

     (use

     *

     or "required"), and explain that array/object-type fields must be passed in via

     --body

     JSON.
4. Business Process (only when there are multi-step links): Describe links that require filling in the response of the previous step (e.g., "check fee → fee calculation" "create coupon → sell coupon → issue coupon"), emphasize that fields must be taken from the previous step's response and cannot be fabricated. If there is no strong dependency chain, write "all commands are independent queries".
5. Examples: Provide 2-4 commands that can be directly copied and run, including at least one read example and (if there are write commands) one write example with

   --yes

   ; try to take parameters from

   sampleBody

   of catalog or test parking lots in the shared base.

• The main body of SKILL.md must be short (resident in context, control tokens). Sink large, low-frequency content to

  references/<topic>.md

  , and use relative links in the main body to indicate "Read when X is needed".
• Content suitable for references: complete field dictionary / long enumeration tables (e.g., full set of coupon types, vehicle types), complete business processes with more than two steps, detailed tables of error codes to disposal actions.
• Clearly write trigger conditions in the main body, for example: "Before handling the complete link of create coupon → sell coupon → issue coupon, first Read

  references/coupon-flow.md

  ", so that the model loads on demand instead of reading all by default.

• Each

  openydt <domain> <use>

  in the command table must exist in reality, and can only come from two categories, choose one to verify:
  • First-class domain commands: Based on real subcommands in

    cmd/gen/<domain>.go

    (generated by codegen), or verify by running

    openydt <domain> --help

    .
  • Catalog interfaces: Based on

    interfaces[]

    marked

    included: true

    in

    catalog/catalog.json

    , each includes

    cmd

    (business code) /

    domain

    /

    readwrite

    /

    params

    /

    sampleBody

    , which can be used to infer command names, read/write attributes and example parameters.
• Do not fabricate command names or parameters. Interfaces marked

  included:false

  (with

  excludeReason

  ) are out of scope, do not create Skills for them.
• If the target interface does not have first-class domain commands, you can use the general fallback

  openydt api <cmd> --body '{...}'

  to call it in the Skill (see the three-layer command model of shared), but still confirm that the

  cmd

  exists in the catalog and is callable.

• Operations marked

  readwrite: "write"

  in the catalog (or any operation that changes the platform state: pay fee, open gate, issue coupon, activate monthly pass, add/remove from blacklist, set parking space, etc.) are write commands.
• Mark write commands as "Write (requires

  --yes

  )" in the "Read/Write" column of the command table, and actually add

  --yes

  in examples; it is recommended to preview the signature request with

  --dry-run

  first if necessary. This is consistent with the security rules of shared, and new Skills must not weaken this requirement.

1. Determine the domain/scenario and target interface, verify real commands, read/write attributes, required parameters, and

   sampleBody

   in

   catalog/catalog.json

   (

   included:true

   ) or via

   openydt <domain> --help

   .
2. Create

   skills/openydt-<name>/SKILL.md

   , fill in according to the above frontmatter and main content skeleton; enrich Chinese trigger words in description.
3. Verify the authenticity and read/write marking of each command in the command table, mark

   --yes

   for write commands.
4. Sink large content to

   references/

   , leave on-demand loading guidance in the main body.
5. Self-check: whether all commands exist in reality, whether all write operations are marked with

   --yes

   , whether it requires reading shared at the beginning, whether the description has sufficient trigger words.

CRITICAL：开始前 MUST 先用 Read 工具读取

../openydt-shared/SKILL.md

（认证 / profile / 签名 / 状态码 / 限速 / 安全规则）。未读共享基座不要执行任何命令。

• <这类诉求> → 留在本域。
• <那类诉求> → 改用

  openydt <其它域> --help

  。

<use>

openydt <域> <use>

--body '<json>'

openydt <域> <use-read>

--xxx

openydt <域> <use-write>

--yes

--xxx

yyyList

--body

标

*

为必填。

1. 读示例：

openydt <域> <use-read> --xxx <值>

1. 写示例（写操作必须加

   --yes

   ，必要时先

   --dry-run

   ）：

openydt <域> <use-write> --yes --body '{"xxx":"...","yyyList":[{...}]}'

undefined