功能描述

Use when the user mentions wos, WOS, WoS, or Web of Science and wants topic-based literature search, Shenzhen University library login, paper screening, abst...

使用说明 (SKILL.md)

WOS Literature To Feishu

Name: aaa
Author: samadhifire

Before executing, read references/playbook.md.

Core Rules

Trigger this skill whenever the user mentions wos, WOS, WoS, or Web of Science.
Before touching WoS or Feishu, ask the preflight questions from the playbook in one batch. If the required inputs are still missing, stop and wait for the user instead of searching immediately.
Prefer the Shenzhen University library route first: https://www.lib.szu.edu.cn/er?key=web+of+science
Prefer Web of Science - SSCI when the user does not specify another database.
Browser automation for WoS is not fixed to one tool. If the user explicitly says they are using playwright-cli, use the playwright-cli skill for browser steps and do not silently switch to opencli-browser. Only use another browser automation path when the user agrees or the requested path is unavailable.
Prefer local lark-cli for Feishu Base writeback. Do not default to browser-based Feishu data entry.
If login reaches Shenzhen University unified auth and asks for SMS code, verification code, captcha, or other second-step authentication, stop and wait for the user to complete it before continuing.
Default SZU username may be prefilled as 2410032057 for this user's local workflow, but keep the password runtime-only and do not persist it into the skill.
Do not store passwords, verification codes, or other credentials inside the skill, repo, or local files. If the user shares credentials in the current chat, treat them as runtime-only and do not persist them into the skill.
Do not silently switch to direct Clarivate login while the SZU route is available. Use direct Clarivate institution login only as a fallback.
Do not write into the wrong Feishu subtable. Confirm the exact Base link and target table handling first.
If the user has no clear screening rule, use the simple default: 高相关 + 高引用 + 近5年代表作 + 摘要完整 + 去重 + 子主题覆盖均衡

Execution Order

Ask the preflight questions from the playbook.
Lock the search scope, target count, and screening rule.
Expand the user's Chinese topic into English core concepts, synonyms, adjacent concepts, and object terms before building final queries.
Enter WoS through the SZU library route and wait at any verification step.
Build query buckets and collect candidate papers.
Deduplicate and screen down to the target count.
Extract at least 标题 / 年份 / 作者 / Q几区 / 引用数 / 摘要 / 文章链接 / 抓取时间 / 主题标签 / 现在状态, and actively resolve Q几区 whenever that field exists in the target schema.
Use local lark-cli to create or update the Feishu table fields and records.
Verify final table name, field structure, sample rows, and record count.

Important Defaults

Database default: WOS Core Collection -> SSCI
Document type default: Article + Review
Language default: English
Count default when the user says "先来一批": 50
Base field default: 标题 / 年份 / 作者 / Q几区 / 引用数 / 摘要 / 文章链接 / 抓取时间 / 主题标签 / 现在状态
Q几区 default semantics: JCR Quartile, and it should be checked by default when that field is present.

Journal Quartile Rule

SSCI几区 / JCR几区 / 中科院分区 is journal-level metadata, not article-level metadata.
If the target table includes Q几区, do not silently skip it during writeback.
Unless the user explicitly asks for 中科院分区, interpret Q几区 as JCR Q1/Q2/Q3/Q4.
Before writing records, either fill Q1/Q2/Q3/Q4 or explicitly report that quartile verification is still pending.
Do not treat Q几区 as optional just because it requires an extra journal-level lookup step.
Abstract extraction is normally feasible from the WoS full record page and can be written into Feishu Base.

Default Field Types

标题: text/plain
年份: number, integer
作者: text/plain
Q几区: select, single choice, options Q1/Q2/Q3/Q4; if the workflow often needs staged completion, add 待查
引用数: number, integer
摘要: text/plain
文章链接: text/url
抓取时间: datetime
主题标签: text/plain
现在状态: select, single choice, options 已读摘要/已读全文, blank allowed

Windows Notes

On Windows PowerShell, if lark-cli resolves to lark-cli.ps1 and is blocked by execution policy, explicitly call lark-cli.cmd.
If the user chooses playwright-cli, prefer playwright-cli snapshot for page inspection, and use a persistent profile or the user's existing session strategy only after confirming it with the user.
When using --json @file in PowerShell, quote the value, for example: --json "@.\\feishu\\record_payloads\\001.json" \r

安全使用建议

This skill appears to do what it says (search WoS and write to Feishu via lark-cli), but there are some mismatches you should resolve before use: 1) The metadata lists no required binaries or env vars, yet the instructions depend on local lark-cli and optionally playwright-cli — ensure those tools are installed and up-to-date. 2) The workflow will require Shenzhen University credentials (or interactive login) and a Feishu base token/session; do not paste passwords or codes into chat and prefer interactive login as the playbook recommends. 3) Confirm you trust the environment that will run the agent (local CLI calls will be executed by the agent if you permit it). 4) If you want a tighter security posture, request that the skill owner add explicit metadata for required binaries and a clear statement of what runtime secrets are needed, or run the workflow manually following the playbook instead of giving the agent permission to execute commands. If you accept these caveats, the skill is functionally coherent; if not, treat it as untrusted.

功能分析

Type: OpenClaw Skill Name: aaa Version: 1.0.0 The skill automates academic literature retrieval from Web of Science via the Shenzhen University portal and exports results to Feishu Base. It utilizes high-risk capabilities including browser automation (playwright-cli), local shell execution (lark-cli), and file system access for record payloads. While the logic is aligned with the stated purpose and includes security-conscious instructions (e.g., waiting for 2FA, forbidding credential storage), it contains hardcoded local environment details such as a specific user path in 'references/playbook.md' and a default student ID in 'SKILL.md', which meet the threshold for a suspicious classification due to the use of risky capabilities and environment-specific data.

能力评估

ℹ Purpose & Capability

The SKILL.md is clearly about searching Web of Science and writing results to Feishu Base via a local lark-cli; that capability aligns with the description. However, the registry metadata does not declare required binaries (e.g., lark-cli, potentially playwright-cli) or any required credentials even though the instructions explicitly rely on them. The skill package name 'aaa' in metadata also doesn't match the internal skill name 'wos-feishu-literature' (minor coherence issue).

✓ Instruction Scope

The runtime instructions are explicit and constrained to the stated task: run preflight questions, enter WoS via the SZU library route, pause for any interactive 2FA, collect and screen records, and write via local lark-cli. The playbook explicitly forbids persisting user passwords and requires pausing when verification is needed, which limits risky behavior. No instructions attempt to read unrelated host files or exfiltrate data to unknown endpoints.

✓ Install Mechanism

This is an instruction-only skill with no install spec or embedded code, which is low-risk from an installation perspective. Nothing is downloaded or written by the skill bundle itself.

⚠ Credentials

The skill declares no required env vars or binaries, yet the playbook assumes presence of local tooling and credentials: lark-cli (with --base-token / auth), possibly playwright-cli and the user's Shenzhen University credentials or session. That mismatch is notable: the skill will require the user's Feishu base token or a logged-in lark-cli session and may expect a SZU login flow. Users should not assume the skill is self-contained; sensitive tokens/credentials will be needed at runtime and are not declared in metadata.

✓ Persistence & Privilege

The skill does not request permanent 'always' inclusion and does not claim to write to other skill configs or system-wide settings. It instructs not to persist passwords and to pause for interactive auth, which reduces privilege concerns. Note: model invocation is enabled (default), which lets the agent call the skill autonomously; this is normal but the user should be aware the agent might attempt operations if given runtime permission.

版本历史

v1.0.0

Initial release of wos-feishu-literature skill for streamlined WoS-to-Feishu Base workflows. - Triggers on user mention of wos, WoS, or Web of Science for literature collection, screening, and export to Feishu multidimensional tables. - Prefers Shenzhen University library login and local lark-cli for data writeback; uses browser automation per user’s explicit tool choice and waits at authentication steps. - Asks all needed preflight questions upfront, confirms search scope, screening rule, and Feishu Base setup before continuing. - Deduplicates and screens papers by default rules, expands topics into English concepts and keywords, and extracts detailed metadata (including JCR Quartile, when required). - Carefully processes credentials as runtime-only, never storing sensitive user info. - Provides field type defaults, workflow-specific handling for Q几区 (quartile) assignment, and platform/Powershell compatibility notes.

元数据

Slug aaa

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题

aaa 是什么？

Use when the user mentions wos, WOS, WoS, or Web of Science and wants topic-based literature search, Shenzhen University library login, paper screening, abst... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 69 次。

如何安装 aaa？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install aaa」即可一键安装，无需额外配置。

aaa 是免费的吗？

是的，aaa 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

aaa 支持哪些平台？

aaa 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了 aaa？

由 SamadhiFire（@samadhifire）开发并维护，当前版本 v1.0.0。

aaa