Token导航 LogoToken导航

Agent Skill · 研究检索

dream(dream 搜索)

dream 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 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。

dream 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为执行命令;许可证信息还不明确,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 58,892。

总安装 399周安装 16GitHub Stars 318下载 129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
unknown

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:dream(dream 搜索)
来源仓库:https://github.com/boshu2/agentops
仓库路径:skills/dream
安装命令:
npx skills add https://github.com/boshu2/agentops --skill dream
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 npx skills 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。

skills.shnpx skills
npx skills add https://github.com/boshu2/agentops --skill dream

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。

dream 提供私有夜间操作员模式,支持本地 overnight 化合物验证与晨会报告解读。区分公共 nightly proof 与本地 lane,保护敏感数据不被外传。可渲染 DreamScape 可视化与 Council 决策过程,辅助人工判断。适用于高风险任务预演,降低生产环境试错成本。

dream 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/dream

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 Codex、Claude、Cursor、Gemini CLI;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 npx skills add https://github.com/boshu2/agentops --skill dream;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

dream 怎么安装?

在终端执行 npx skills add https://github.com/boshu2/agentops --skill dream。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dream 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。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。

dream 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 npx skills add https://github.com/boshu2/agentops --skill dream 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。

dream 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

dream 用的人多吗?

公开数据是累计安装 399 次,近一周安装 16 次,下载 129 次,GitHub 318 星。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。

dream 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

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 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。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 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 Codex,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。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 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
dream

dream 搜索 · 本页

399318执行命令未明确通过

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

Dream - Private Overnight Operator Mode

/dream is the interactive face of the Dream system. It should drive the same engine as ao overnight; do not invent a parallel workflow.

Purpose

Use Dream to:

  • preview or persist Dream setup and scheduler assistance
  • run a bounded private overnight compounding pass against the real .agents/ corpus
  • render and interpret the latest morning report, including Dream Council and DreamScape sections
  • explain the difference between the private local Dream lane and the public nightly proof harness

Routing

Map user intent to one of three lanes:

IntentCommandNotes
bootstrap, install, configure Dreamao overnight setupDefault to preview. Use --apply only when the user explicitly wants config or scheduler artifacts persisted. Detects Tier 1 local curator separately from Tier 2 Dream Council runners.
inspect or feed local Gemma curator`ao overnight curator statusdiagnose
run Dream nowao overnight startInclude --goal when the user provides one. Use --runner and --creative-lane only when the user asks for multimodel or wildcard analysis.
inspect a prior runao overnight reportUse --from <dir-or-summary.json> for non-default report paths.

Compounding Loop (v2)

Dream v2 runs a bounded outer loop of INGEST -> REDUCE -> MEASURE iterations until a halt condition fires: wall-clock budget exhausted, plateau (K sub-epsilon deltas in a row), regression beyond a per-metric floor, or metadata integrity failure. Each iteration is atomic and checkpointed so any rollback leaves the corpus clean. Dream is strictly knowledge-only.

Anti-goals (hard constraints):

  • NEVER mutates source code.
  • NEVER invokes /rpi or any code-mutating flow.
  • NEVER performs git operations (no commits, branches, push, rebase, checkout).
  • NEVER creates symlinks anywhere.
  • No swarm/gc agent fan-out inside iterations. Bounded in-process read-only finding generators may run concurrently during INGEST, but REDUCE remains the single serialized writer for .agents/rpi/next-work.jsonl.

Tier 1 Curator And Trigger Mesh

Dream can expose a local Tier 1 curator through dream.local_curator.*. The first supported shape is Ollama + Gemma, backed by an operator-owned worker directory such as D:\dream and a vault such as D:\vault. This is not a Dream Council runner: Gemma drafts, lints, triages, and writes auditable queue or event records; Codex and Claude remain Tier 2 review/synthesis runners; humans own promotion into durable authored memory.

Use ao overnight curator status --json to check the worker, queue, model, and Ollama endpoint. Use ao overnight curator enqueue --kind lint-wiki|dream-seed or --kind ingest-claude-session --source <path> --chunk-start <n> --chunk-end <n> only for allowlisted knowledge jobs. Use ao overnight curator event when Gemma or a local SOC signal needs Tier 2 attention. Events carry source, severity, desired action, escalation target, and budget; no runner should recursively invoke another runner without consuming an explicit event budget and leaving a ledger entry.

Key Rules

  • Keep the shared control plane authoritative. Dream settings live under dream.*; do not duplicate config logic inside the skill.
  • Keep platform semantics honest. Never promise scheduled execution on a sleeping laptop.
  • Keep the first slice bounded. Dream is for close-loop, defrag preview, metrics, retrieval proof, optional briefing, and optional artifact-mediated council synthesis.
  • Do not imply tracked source-code edits overnight unless the runtime actually supports them.
  • GitHub nightly is the public proof harness. Dream is the private local engine.

Execution Steps

Step 1: Resolve the operator lane

Interpret the request as setup, start, or report.

Examples:

/dream setup
/dream setup --apply --runner codex --runner claude --at 01:30
/dream start "close the loop on today's auth work"
/dream report
/dream report --from .agents/overnight/latest

Step 2: Setup lane

Use ao overnight setup to inspect the host, available runtimes, scheduler mode, and keep-awake behavior.

ao overnight setup
ao overnight setup --apply --runner codex --runner claude --at 01:30

Default behavior:

  • Preview first when the user asks how Dream would work on this machine.
  • Persist with --apply only when the user explicitly asks to save config or generate scheduler artifacts.

Expected outputs:

  • a config preview in terminal, JSON, or YAML
  • optional generated scheduler assistance under .agentops/generated/dream/
  • optional dream.local_curator config when a supported local curator is configured or detected

Step 2a: Local curator lane

Use ao overnight curator when the user asks about Gemma, the local worker, the SOC trigger path, or Tier 1 drafts.

ao overnight curator status --json
ao overnight curator diagnose
ao overnight curator enqueue --kind lint-wiki
ao overnight curator enqueue --kind dream-seed
ao overnight curator compact --dry-run
ao overnight curator event --source local-soc --severity high --desired-action "review alert cluster" --budget 1

Do not promote Tier 1 drafts directly into authored content. Treat draft promotion as Tier 2 or human review work.

Step 3: Bedtime run lane

Use ao overnight start for the actual local run.

ao overnight start --goal "close the loop on today's auth work"
ao overnight start --goal "stabilize release follow-ups" --runner codex --runner claude --creative-lane
/dream start --queue=.agents/dream/tonight.md
/dream start --max-iterations=3
/dream start --warn-only=false

Expected behavior:

  • operates against the real repo-local .agents corpus
  • writes summary.json and summary.md
  • degrades honestly when soft-fail steps or keep-awake helpers are unavailable

Step 4: Morning report lane

Use ao overnight report to render the latest Dream result.

ao overnight report
ao overnight report --from .agents/overnight/latest
ao overnight report --from .agents/overnight/latest/summary.json

Focus the response on four questions:

  1. What state did I wake up to?
  2. What ran overnight?
  3. What degraded or failed?
  4. What should I do first?

Output

  • Dream setup preview or persisted dream.* config with optional scheduler assistance artifacts
  • Dream morning report artifacts:

- .agents/overnight/<run>/summary.json - .agents/overnight/<run>/summary.md

  • A concise operator summary with degraded items and the single highest-signal next action
  • v2 morning reports also include per-iteration sub-summaries, composite fitness_delta, and plateau_reason / regression_reason when applicable

Examples

Help me get Dream working on this Mac without pretending launchd survives sleep.
Run Dream tonight with Codex and Claude, goal: stabilize the Homebrew release follow-up.
Read the latest Dream report and tell me the first move.

Reference Documents

Troubleshooting

ProblemCauseFix
ao overnight setup suggests manual scheduler modeHost scheduler semantics are ambiguous or unsupportedKeep the run operator-armed, or use generated assistance only after reviewing the host behavior
Dream report shows degraded keep-awakecaffeinate or the platform helper is unavailableContinue with the degraded report; do not claim the machine will stay awake
No morning report existsDream has not been run yet or the output dir is differentRun ao overnight start, or point ao overnight report --from at the correct directory
User expects GitHub nightly to replace local DreamCI proof harness and local Dream are different surfacesExplain that nightly proves the contract, while Dream operates on the private local corpus

Delineation vs /evolve

LaneRunsMutates code?Mutates corpus?Outer loop?Budget
/dreamnightly, private localNoYes (heavy)Yes (convergence)wall-clock + plateau
/evolvedaytime, operator-drivenYes (via /rpi)Yes (light)Yescycle cap

Dream owns the knowledge compounding layer; /evolve owns the code compounding layer. Both share fitness-measurement substrate via corpus.Compute / ao goals measure. Run Dream overnight, then start each day with /evolve against the freshly-compounded corpus with a clean fitness baseline.

See Also

  • /handoff - capture explicit session closeout before bedtime
  • /status - inspect the current repo state before deciding whether to run Dream
  • /compile - compile the knowledge corpus when the morning report points at corpus hygiene work
  • /rpi - run the daytime delivery flow before handing work off to Dream

平台分布

Codex

37.83%
按下载量换算49

Claude

30.52%
按下载量换算39

Cursor

20.64%
按下载量换算27

Gemini CLI

8.48%
按下载量换算11

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

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 npx skills add https://github.com/boshu2/agentops --skill dream 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