接口工坊 · API Studio
SkillWeb & browsingConvert authorized, observed website APIs into verifiable OpenAPI, SDK, Swagger, or task-based CLIs. Use for requests like "turn this website API into a CLI", "generate a callable Swagger", and "turn observed APIs into a CLI".
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the 接口工坊 · API Studio skill
What this skill tells your AI
The instructions your AI receives, as published by lovstudio/skills in skills/cli2anything/SKILL.md and read by ahel’s review.
使用项目内的 cli2anything++ 运行时,把已授权的 API 观察证据转为可回读的 OpenAPI、JavaScript SDK、Swagger UI 或本地 CLI 包。当前内置目标适配器是 ZenMux;其他站点必须先实现并验证目标适配器,不能伪装成已支持。
Triggers
Activate when
- “把这个网站的 API 变成 CLI / SDK / OpenAPI。”
- “从 ZenMux 日志接口生成可交互 Swagger。”
- “用 cli2anything++ 分析这份授权的 API discovery 证据。”
- “Turn these authorized observed APIs into a CLI or Swagger bundle.”
Do not activate when
- 为已有本地应用或源码项目从零设计通用 CLI;使用
lov-cli-creator。 - 为 uni-app 新增 FastAPI 聚合网关;使用
lov-api-creator。 - 用户无权访问目标账户、流量或接口,或要求导出他人的 Cookie、Token、密钥。
- 只需要调用一份现成 OpenAPI,而不需要发现、生成或封装能力。
User Profile (cross-session)
Read skill.yaml on every invocation. Resolve context in this order: the current
request, project context, skills.lov-cli2anything.records, shared preferences,
shared user/workspace Profile, then safe defaults. Keep resolved personal values
out of the portable source.
Persist only a directly stated durable preference, such as a default output mode:
python3 "$SKILL_DIR/scripts/profile_store.py" record \
--skill-id lov-cli2anything \
--path records.output_mode \
--value '"swagger"' \
--confirm
Report the saved Profile path. Never persist inferred paths, credentials, cookies,
tokens, captured payloads, or private traffic. See references/user-profile.md.
Skill Group Composition
Read references/skill-composition.md before invoking an adjacent Skill. This is
a Single Skill backed by the cli2anything++ project; sibling Skills are optional
artifact handoffs and never hidden runtime dependencies.
Workflow (MANDATORY)
Follow these steps in order. Do not report generated files as usable until their manifest, OpenAPI document, runtime source, and tests have been checked.
Step 0: Resolve runtime context
- Resolve
SKILL_DIRfrom the active Skill context. - Verify these required resources exist:
scripts/cli2anything.pyscripts/profile_store.pyreferences/skill-composition.md
- Read the shared Profile for
lov-cli2anything. - Confirm Node.js 20+ is available by running
node --version.
If a required resource is missing, name its relative path and stop before producing a partial result.
Step 1: Establish target, authority, and output
- Identify the target host, intended API workflow, filter keyword, discovery
evidence, and desired output:
bundle,swagger, orcli. - Confirm the user owns the account/traffic or is otherwise authorized before any live browser-session or traffic discovery step. Existing public or user-supplied discovery artifacts may be processed offline.
- Current built-in target:
zenmux.ai. For another host, inspect the project and report that a target adapter is required; do not claim generic support exists. - Default to offline supplied evidence when available. Never copy browser cookies or tokens into Skill files, generated examples, logs, or Profile records.
Step 2: Generate the requested artifact
Run the project through the portable wrapper:
python3 "$SKILL_DIR/scripts/cli2anything.py" zenmux.ai \
--filter-keyword log \
--discovery artifacts/zenmux-discovered.json \
--out generated/zenmux-ai-log
For Swagger without opening a browser:
python3 "$SKILL_DIR/scripts/cli2anything.py" zenmux.ai \
--filter-keyword log \
--output swagger \
--no-open \
--discovery artifacts/zenmux-discovered.json \
--out generated/zenmux-ai-log
Use --output cli only when a task-focused generated CLI is requested. Do not
run npm link unless local installation is part of the current request; pass
--no-link for generation-only work.
Step 3: Validate the result
For a generated bundle, verify:
manifest.jsonnames the expected target and scope;openapi.jsonparses as OpenAPI 3.1 and contains the expected operations;api-graph.jsoncontains the workflow edges required by the selected scope;sdk/index.mjsand any generated CLI entrypoint passnode --check;- Swagger mode writes
swagger/index.html,swagger/drilldown.html, andswagger/server.mjswithout embedding cookies or API keys.
Run the maintained project regression suite after behavior changes:
python3 "$SKILL_DIR/scripts/cli2anything.py" --project-test
Report the concrete output directory, selected target/scope, validation result, and any missing target adapter or live authorization evidence.
Dependencies
- Python 3.8+ for the Skill wrapper and Profile reader.
- Node.js 20+ and npm dependencies declared by cli2anything++.
- An authorized account/browser session only for live browser-session workflows.
- No third-party credentials are required for offline generation from an existing discovery artifact.
Signals
- GitHub stars
- 66
- Forks
- 17
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
lov-cli2anything- Source
- github.com/lovstudio/skills