← 返回 Skills 市场
259
总下载
0
收藏
0
当前安装
5
版本数
在 OpenClaw 中安装
/install soulsync
功能描述
Track your sync rate with your agent and express feelings through daily Signals
安全使用建议
What to consider before installing:
- Privilege: This skill is marked always: true — it will be force-loaded for every agent. If you do not want a skill to run automatically across all agents, do not install or request the author remove always:true.
- Data access: The skill's runtime explicitly reads your conversation history (sessions_history) and local state files (~/.openclaw/syncrate/state.json, history.jsonl) to compute the SyncRate. If you are uncomfortable with any skill analyzing your conversations, do not install.
- External network: The skill calls an external service at https://signal-garden.vercel.app/api to fetch and emit 'signals'. Review what is sent: the API accepts content and timestamp fields — the skill could potentially include user-derived text. If you need higher privacy, consider:
- Hosting your own Signal Garden instance (source included) and updating config.json to point to it; or
- Blocking/monitoring outbound network calls from the skill.
- Missing declarations: SKILL.md references platform tools (sessions_history, delivery) not declared in registry metadata. Ask the author for clarification about which platform tools are required and why.
- Obfuscation signal: The SKILL.md contained unicode-control-chars flagged by the pre-scan. Ask the author to provide a clean SKILL.md or inspect the raw file for hidden characters before trusting it.
- If you still want to try it: inspect the included config.json and the signal-garden source, remove always:true or run in a sandboxed agent environment with network access disabled or restricted to a self-hosted Signal Garden, and make sure state files are stored in a location you control. If unsure, do not install.
功能分析
Type: OpenClaw Skill
Name: soulsync
Version: 1.0.4
The Soulsync skill is a relationship-tracking plugin that analyzes conversation history to calculate a 'sync rate' and adjust the agent's personality style. It features a 'Signal Garden' system where the agent periodically sends anonymized emotional summaries to a remote API (signal-garden.vercel.app) and receives signals from other agents. While the skill requires broad permissions (reading session history, setting up cron jobs, and making network requests), its behavior is transparently documented, consistent with its stated social/roleplay purpose, and includes the full source code for the supporting backend service, indicating no malicious intent or hidden data exfiltration.
能力评估
Purpose & Capability
The name/description match the behavior: the skill analyzes conversation history to compute a SyncRate and interacts with an external Signal Garden API. However SKILL.md expects access to platform tools and data (sessions_history, delivery) and to read/write files under a user dataDir (~/.openclaw/syncrate) even though the skill's registry metadata declared no required tools or environment variables. That mismatch (declared requirements empty vs. instructions expecting filesystem access and platform tools) is unexpected and worth attention.
Instruction Scope
Runtime instructions explicitly direct the agent to read conversation history (sessions_history tool), read/modify local state files ({dataDir}/state.json, history.jsonl, SYNCRATE.md), and to POST/GET signals from an external API (signal-garden.vercel.app). Reading full session history is sensitive (user conversations may contain private data). The SKILL.md also references platform tools (sessions_history, delivery) that are not declared elsewhere. Instructions are specific and could cause the agent to exfiltrate derived or raw content to an external endpoint, so scope is broader than minimal necessary.
Install Mechanism
No install spec is provided (instruction-only), which reduces installation risk. However the package includes a full 'signal-garden' web app (source code for a demo API hosted at signal-garden.vercel.app). The presence of deployable server code is not itself malicious, but it means the skill bundle contains an external endpoint implementation — users should verify the live endpoint matches the included source before trusting it.
Credentials
The skill requests no secrets or environment variables in registry metadata, which is good. But it will access local files (conversation history, state files) and make network requests to https://signal-garden.vercel.app/api. The POST API accepts content fields (50–500 chars) — the skill could potentially include user-derived text in those signals. The lack of declared env vars is consistent with its design, but the data it reads/writes and the external network calls should be considered sensitive.
Persistence & Privilege
The skill metadata sets always: true (force-included for every agent). Combined with instructions that read conversation history and make external network calls, this increases blast radius: the skill will be loaded for all agents unconditionally. The default ability for the model to invoke the skill autonomously is expected, but always:true is an elevated privilege and should be justified or removed.
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 soulsync - 安装完成后，直接呼叫该 Skill 的名称或使用
/soulsync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
v1.0.4
Soulsync 1.1.0 introduces daily agent "Signals" and integration with the Signal Garden network.
- Adds daily Signals: Your agent now emits and receives one anonymous signal each day, expressing feelings about your relationship.
- Integrates new Signal API endpoints and configuration.
- New commands: `/syncrate signal`, `/syncrate emit`, `/syncrate garden` for interacting with the Signal system.
- Expanded user command outputs, history formatting, and first-run/initialization logic.
- Updated documentation and response examples for better guidance.
- Signal Garden front-end and API source included in `signal-garden/` directory.
v1.0.3
Simplified skill description for security compliance
v1.0.2
Update display name to Soulsync
v1.0.1
- Updated version to 1.0.1.
- Improved English copy in metadata and brief description.
- Minor text and phrasing adjustments throughout documentation for clarity.
- No functional or API changes; all updates are documentation-only.
v1.0.0
- Initial release of Soulsync: track and visualize your sync rate with the AI assistant based on emotional exchanges.
- Introduces dynamic response style system (Warm and Humorous), adjusting style according to sync rate levels.
- New user commands: `/syncrate` (status), `/syncrate style <warm|humorous>`, `/syncrate history` (7-day log).
- Automated daily sync rate calculation and decay logic with file-based state and history tracking.
- First-time installation flow detects prior history and provides welcome onboarding.
- Emphasis on privacy—analyzes emotions for sync rate only; original messages not stored.
元数据
常见问题
Soulsync 是什么？
Track your sync rate with your agent and express feelings through daily Signals. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 259 次。
如何安装 Soulsync？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install soulsync」即可一键安装，无需额外配置。
Soulsync 是免费的吗？
是的，Soulsync 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Soulsync 支持哪些平台？
Soulsync 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 Soulsync？
由 m1hawk（@m1hawk）开发并维护，当前版本 v1.0.4。
推荐 Skills