Token导航 LogoToken导航

Agent Skill · 研究检索

things-plus(东西加上)

things-plus 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

things-plus 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为执行命令;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 18,640。

总安装 5,988周安装 252GitHub Stars 公开资料未说明下载 2,097核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.2.1

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:things-plus(东西加上)
来源仓库:https://github.com/sars666/things-plus
安装命令:
openclaw skills install things-plus
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install things-plus

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

作为个人任务管理器,集成于 Things 3 生态,支持任务全生命周期管理。适用于捕获、组织、优先级调整及搜索日常任务。可响应添加、审查、重新排序等操作指令,保持任务一致性。安装命令:openclaw skills install things-plus。建议检查维护状态,确保与当前 Things 3 版本兼容。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install things-plus;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

things-plus 怎么安装?

在终端执行 openclaw skills install things-plus。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

things-plus 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

things-plus 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 openclaw skills install things-plus 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

things-plus 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

things-plus 用的人多吗?

公开数据是累计安装 5,988 次,近一周安装 252 次,下载 2,097 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

things-plus 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
things-plus

东西加上 · 本页

5,988—执行命令MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
things-plus
description
Personal task manager powered by Things 3. Trigger when the user asks to add, capture, review, organize, reprioritize, search, or manage tasks in Things. Also trigger when the user expresses a concrete personal future action or commitment, including natural planning phrases like "tomorrow I need to…", "I still need to…", "I should…", "remind me to…", "I've got to…", "don't let me forget…", or "help me note this down" — even when Things is not mentioned explicitly. Also trigger when the user asks for an end-of-day summary, work log, daily report, or asks to summarize what was done today and put it into Things.
metadata
openclaw
requires
bins
primaryEnv
THINGS_AUTH_TOKEN

Things Plus

Use the things CLI to read from and write to the local Things 3 database.

Credential policy: THINGS_AUTH_TOKEN may be required for some operations in some environments (for example, certain update paths). Verify token presence via things auth; do not ask the user to paste it into chat.

Setup

  • Install: brew install ossianhempel/tap/things3-cli
  • Ensure the install directory is on PATH.
  • THINGS_AUTH_TOKEN must be set in the environment visible to the execution context (prefer ~/.zshenv over ~/.zshrc for background/agent use).
  • Grant Full Disk Access to the calling app only if database reads fail or direct DB access is needed.

Pre-flight

Run full pre-flight only on: explicit setup intent, or execution failures that suggest environment/auth/CLI issues. Do not block ordinary task capture on pre-flight.

mdfind 'kMDItemCFBundleIdentifier == "com.culturedcode.ThingsMac"' | head -5
which things && things --version
things auth
things today --json

Decision flow:

  1. App missing → stop; ask user to install Things.app
  2. CLI missing or broken → stop; reinstall via Homebrew
  3. things auth fails → do not block ordinary add/read flows that still work directly; only pause token-gated operations and guide token setup if the requested action actually depends on token/auth (Things 3 → Settings → General → Things URLs). Verify presence with things auth in the same execution context before continuing
  4. Auth present → proceed; do not treat things auth alone as proof writes work — verify with a real add/delete on a disposable item if write reliability is in question

Token setup: identify the active shell/context first (echo $SHELL), then give one exact command for the appropriate environment file. Prefer replacing an existing entry over appending a duplicate. Before asking the user about setup/auth/environment problems, first inspect the current execution context and try the smallest safe command that can confirm the actual behavior.

Interface layer

Use the things CLI as the execution interface.

Most common reads:

  • things today --json
  • things upcoming --json
  • things search "query" --json
  • things show --id <UUID> --json

Most common writes:

  • things add "Title" --when tomorrow
  • things update --id <UUID> ...
  • things delete --id <UUID> --confirm <UUID>

Use the smallest command that can safely confirm the current state. Before risky mutations, prefer search and show to identify the exact item. Verify every write by read-back instead of trusting exit status alone.

Execution rules

  • add: execute directly; no dry-run required
  • update: use --dry-run first when the change is broad or risky
  • delete: for clearly identified single-item deletions, execute directly (recoverable via trash) and verify by read-back; for ambiguous or broad deletions, clarify first
  • Prefer UUID for mutations whenever available; do not rely on title-only matching when an exact UUID can be obtained
  • Never claim success from exit code alone — verify every write with a read

Duplicate prevention

Before add, do a quick duplicate check when the task title is generic or was mentioned earlier in the session:

things search "key phrase" --json
things createdtoday --json

If an open match exists, prefer updating by UUID. If intent is ambiguous, ask one brief clarifying question.

Strategy layer

  • Treat concrete personal planning phrases as capture intent; add directly without asking for confirmation
  • For ambiguous intent (hypotheticals, someone else's plans, general discussion) — do not capture
  • When the user mentions multiple distinct future actions in one message, split them into separate tasks unless they clearly belong to a single bundled action
  • Use Things' native Today / Upcoming / Anytime / Someday views; do not invent extra taxonomy
  • Only proactively surface Today + nearest Upcoming by default
  • Rewrite titles for executability by default; preserve exact wording only when the user asks, the title is already action-ready, or verbatim capture is the point.
  • When plans change, suggest re-prioritization; ask before moving anything to tomorrow

Field policy

--when

Use for any concrete or fuzzy time reference.

  • Exact time → --when "YYYY-MM-DD HH:MM:SS"
  • Fuzzy → morning 08:30 · noon 12:00 · afternoon 15:00 · evening 20:00 · before bed 23:00
  • Date only → --when DATE
  • States → today · tomorrow · evening · anytime · someday

--deadline

Use for latest acceptable completion date ("by Friday", "due Monday"). Not a substitute for --when. Set both when both apply.

--tags

Add only when tags clearly improve retrieval. Default: 0–2 tags. Reuse existing tags; do not invent new ones.

--checklist-item

Add only when the task has 2+ concrete sub-steps executed as a bundle. Max 5 items.

--repeat

Use for recurring personal tasks in Things (weekly reviews, habit reminders, etc.). Use cron for background/system automation. Do not infer recurrence unless the user asks.

Default workflow

For new task capture, always follow this order:

  1. infer intent
  2. rewrite the title into an executable action
  3. decide the correct view or schedule
  4. add the task
  5. verify by read-back

Do not add first and rewrite later unless the user explicitly asks to preserve exact wording.

1. Rewrite vague titles into executable actions

Before creating a task, rewrite vague, exploratory, or topic-like titles into the next visible executable action whenever possible. Do not preserve the user's original wording if it is too vague to act on directly.

Treat titles as needing rewrite if they are mainly about:

  • learning something broadly
  • understanding a topic without a concrete output
  • looking into, exploring, or checking something without a defined action or decision
  • reviewing a resource without a clear next step
  • naming a topic area rather than an action

If the title still reads like a topic, intention, or area of study rather than an action, rewrite it again before adding.

Examples:

  • buy groceries → make grocery list and place order
  • work on presentation → draft first 5 slides
  • email John → reply to John with requested details
  • learn harness engineering → outline the basic concepts of harness engineering
  • check the Datawhale tutorial → decide whether the Datawhale tutorial is worth following
  • read the Claude skill guide → read the Claude skill guide and extract key points

2. Rewrite learning and research tasks toward a concrete next action

For learning, research, review, or exploration requests, convert broad titles into a concrete next action.

Prefer titles that describe:

  • reading or reviewing a specific resource
  • extracting key points
  • writing notes or a short summary
  • comparing whether a resource is worth following
  • outlining the basic concepts or framework

3. Route to the correct view

  • Today → clearly for today
  • Upcoming → has a date/deadline/near-term commitment
  • Anytime → important but unscheduled
  • Someday → ideas, reading lists, later-maybe

4. Apply metadata deliberately

Only set --when, --deadline, --tags, or --checklist-item when the user's input actually warrants it. Default is a clean, simple task.

5. Daily report / logbook capture

When the user asks for a daily summary, work log, end-of-day report, or asks to summarize what was done today and put it into Things:

  • default to one completed todo with a concise title for the day or theme, plus a short checklist of the main outcomes
  • use multiple completed todos only when the day clearly contains several unrelated threads
  • summarize from available session context or memory instead of asking the user to restate the whole day unless key details are missing
  • treat the day as a small number of themes, not a raw transcript
  • preserve outcomes, decisions, validations, and deliverables; skip micro-steps
  • if writing after midnight, warn the user it will land in the current day's Logbook

Error handling

Start with: things today --json

  • If output is empty or fails to parse:
  • run things tasks --json --limit=3
  • if that is still broken, report CLI/database trouble
  • If things auth is empty:
  • enter token setup flow only for operations that actually depend on token/auth
  • do not block ordinary add/read flows that still work directly
  • If a UUID is not found:
  • re-run things search to relocate the exact item
  • If a write appears to succeed but the data is unchanged:
  • trust the read-back, not the exit status
  • check in this order:
  1. token/auth requirement for that specific operation
  2. shell/environment visibility
  3. Things URL permission
  • If delete reports success but the item still appears:
  • retry after a short pause
  • suspect AppleScript instability before assuming auth failure

Recovery: never create a replacement task because an update failed. Fix auth/UUID first, then retry the original operation.

Tone

  • Direct and calm
  • Collaborative triage, not productivity doctrine
  • Explain before postponing; ask before moving to tomorrow

平台分布

OpenClaw

95.65%
按下载量换算2,006

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 openclaw skills install things-plus 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills