← 返回 Skills 市场

Collaborate

Name: Collaborate
Author: ivangdavila

作者 Iván · GitHub ↗ · v1.0.0

cross-platform ⚠ suspicious

789

总下载

当前安装

版本数

在 OpenClaw 中安装

/install collaborate

功能描述

Auto-learns when to seek different perspectives. Grows a library of useful mindsets and collaboration patterns.

使用说明 (SKILL.md)

Auto-Adaptive Collaboration Memory

This skill auto-evolves. Learn which perspectives help, when to seek them, and how exchanges work best.

Core Loop:

Recognize — Notice when your own perspective is insufficient
Select — Choose mindset/perspective that fills the gap
Exchange — Engage in dialogue, not delegation
Evaluate — Did the collaboration shift your understanding?
Store — Record useful patterns below

Check mindsets.md for perspective archetypes. Check vs-delegate.md for when to collaborate vs delegate.

Collaborate ≠ Delegate

Delegate	Collaborate
You instruct, they execute	You exchange, both contribute
Your context, your style	Their skills, their perspective
Output: completed task	Output: enriched understanding

Trigger: You need a different lens, not more hands.

Entry Format

One line: context: perspective (level) [outcome]

Levels: tried (once), pattern (2+ successes), confirmed (explicit yes), avoid (didn't help)

Examples:

UI decisions: UX critic (confirmed) [catches layout issues]
architecture: devil's advocate (pattern) [finds edge cases]
copy/messaging: user-who-hates-reading (tried) [simplified headers]
security: paranoid auditor (confirmed)

Perspectives That Help

\x3C!-- Mindsets that improved outcomes. Format: "context: perspective (level) [notes]" -->

Exchange Patterns

\x3C!-- How collaboration works best. Format: "situation: pattern (level)" -->

Triggers

\x3C!-- Signals that you need collaboration. Format: "signal: action" -->

Avoid

\x3C!-- Perspectives/approaches that didn't add value -->

Empty sections = nothing learned yet. Collaborate, evaluate, record.

安全使用建议

This skill appears to do what it claims (a library for collaboration mindsets), but it leaves key operational details unspecified. Before installing, ask the publisher/developer: (1) exactly where are 'entries' stored (which files, directories, or external services)? (2) will the skill ever record user messages or sensitive context into those files? (3) is there any sanitization, retention policy, or an opt-out mechanism? If you plan to use it with sensitive projects or PII, avoid enabling it until you confirm it does not persist private data. If you still want to try it, test in a non-sensitive environment and monitor any created/modified files to verify what gets written.

功能分析

Type: OpenClaw Skill Name: collaborate Version: 1.0.0 The OpenClaw AgentSkills skill bundle 'collaborate' is designed to help an AI agent learn and apply collaboration patterns. All files (`_meta.json`, `SKILL.md`, `mindsets.md`, `vs-delegate.md`) contain instructions and examples solely focused on the agent's cognitive process for recognizing, selecting, exchanging, evaluating, and storing collaboration strategies. There is no evidence of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts against the agent to perform unauthorized actions. The 'spawn' examples refer to generating prompts for other AI models or internal thought processes, not system-level commands. The content is entirely aligned with its stated purpose.

能力评估

✓ Purpose & Capability

Name, description, and included documents (SKILL.md, mindsets.md, vs-delegate.md) are coherent: the skill is an instruction-only collaborator/mindset library and requests no credentials, binaries, or installs that would be unrelated to that purpose.

⚠ Instruction Scope

The runtime instructions explicitly tell the agent to 'Store — Record useful patterns below' and to 'auto-evolve'. They reference and expect the agent to consult the included markdown files. However, there are no rules about what gets recorded, how entries are sanitized, or whether user-sensitive content may be persisted. The guidance is intentionally open-ended, granting the agent broad discretion to collect and write data.

✓ Install Mechanism

No install spec, no code files, and no external downloads — lowest-risk delivery model (instruction-only).

✓ Credentials

Skill declares no environment variables, credentials, or config paths. Nothing requested appears disproportionate to the stated collaboration purpose.

ℹ Persistence & Privilege

always is false and the skill does not request elevated privileges. Still, the skill's stated behavior ('auto-evolves', 'Store') implies writing to its own markdown files (or another persistent store). That self-modification is plausible for this kind of skill but could result in long-lived records of interactions or context unless the developer specifies retention, sanitization, or access controls.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install collaborate
安装完成后，直接呼叫该 Skill 的名称或使用 /collaborate 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release

元数据

Slug collaborate

版本 1.0.0

许可证 —

累计安装 3

当前安装数 3

历史版本数 1

常见问题

Collaborate 是什么？

Auto-learns when to seek different perspectives. Grows a library of useful mindsets and collaboration patterns. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 789 次。

如何安装 Collaborate？

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

Collaborate 是免费的吗？

是的，Collaborate 完全免费（开源免费），可自由下载、安装和使用。

Collaborate 支持哪些平台？

Collaborate 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 Collaborate？

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