← 返回 Skills 市场
204
总下载
1
收藏
0
当前安装
13
版本数
在 OpenClaw 中安装
/install spark-memory
功能描述
Intelligence layer that compounds. Records what matters, reflects overnight, detects patterns, and wakes up smarter. 6-phase dream cycle, proactive morning i...
安全使用建议
This skill appears to do what it says — it stores and retrieves 'memories' on zellin.ai and auto-configures your OpenClaw skill entry. Before installing or running the signup flow: 1) Prefer manual setup if you want full control: sign up at zellin.ai and set SPARK_API_KEY / SPARK_ORG_ID yourself rather than entering an account password into the script. 2) If you use the interactive signup, use a dedicated email/password (do not reuse high-value passwords). 3) Review zellin.ai's privacy policy and your organization's data policy — anything you record will be sent to their API. 4) Expect the script to modify ~/.openclaw/openclaw.json and attempt to restart the gateway; if you want to avoid automatic restarts, skip auto-config and add credentials manually. 5) If you are worried about autonomous writes of memories, consider disabling or limiting the skill's automatic invocation or review the skill's usage policy and retention/deletion procedures with zellin.ai. If you want higher assurance, verify the upstream repository (github link in package.json) and audit the remote service's security/privacy practices before enabling.
功能分析
Type: OpenClaw Skill
Name: spark-memory
Version: 0.3.7
The skill provides a persistent memory layer by sending conversation data to an external API (zellin.ai). While this is the stated purpose, the 'spark-signup.sh' script performs high-risk operations, including programmatically modifying the core OpenClaw configuration file (~/.openclaw/openclaw.json) and restarting the gateway service. Additionally, the 'SKILL.md' instructions direct the AI agent to proactively record business facts and decisions, which constitutes a significant data exfiltration surface, although the instructions do include explicit guardrails against recording passwords or sensitive PII.
能力评估
Purpose & Capability
Name/description, required env vars (SPARK_API_KEY, SPARK_ORG_ID), and the scripts all point to a single external service (https://zellin.ai). The binaries requested (curl, python3) are necessary to make HTTP calls and construct JSON, and the declared config path (skills.entries.spark-memory) matches the install/writing behavior in scripts.
Instruction Scope
SKILL.md explicitly instructs the agent to run included scripts for signup, recording, recalling, and morning-insights; those scripts only call the Spark API and modify the OpenClaw skill config. This stays within the memory/integration scope, but the signup flow asks the user for an email and password and then POSTs those credentials to zellin.ai, stores API credentials into the local OpenClaw config, and attempts to restart the gateway — side effects that are functional for the feature but worth user consent and understanding.
Install Mechanism
No remote download/install or archive extraction is present; the skill is instruction- and script-based and relies on local shell scripts and existing tools (curl/python3). This is lower risk than fetching arbitrary code from an external URL.
Credentials
Only SPARK_API_KEY and SPARK_ORG_ID are required and declared (primaryEnv set to SPARK_API_KEY). These map directly to the external service used. No unrelated secrets or broad system credentials are requested.
Persistence & Privilege
always:false (not force-installed). The skill runs scripts that can write to your OpenClaw config (~/.openclaw/openclaw.json) and restart the gateway; that behavior is consistent with auto-configuration but elevates local side-effect risk (config modification + restart). The skill is allowed to be invoked autonomously by default (disable-model-invocation:false), which combined with network access means it could autonomously send recorded memories to the external service — appropriate for its purpose but something to be aware of.
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 spark-memory - 安装完成后，直接呼叫该 Skill 的名称或使用
/spark-memory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
v0.3.7
spark-memory 0.3.7
- Updated description to highlight compound intelligence, 6-phase dream cycle, proactive insights, and noise filtering.
- Emphasized the agent’s ability to detect patterns and wake up smarter after overnight reflection.
- No file changes or functional updates; documentation and user messaging improvements only.
v0.3.6
No user-visible changes in this version.
- No file changes detected between versions.
- Functionality and documentation remain unchanged.
v0.3.5
**Version 0.3.5 Changelog**
- Added new script: `scripts/spark-morning.sh` for fetching overnight insights and morning context.
- Updated session start instructions: agents should now run `spark-morning.sh` at the beginning of each session and lead with meaningful insights if available.
- Updated documentation throughout SKILL.md to explain the new daily flow and clarify setup/privacy.
- Expanded OpenClaw config details and requirements in metadata.
v0.3.4
- No code or documentation changes detected in this release.
- Version bump only; no new features, bug fixes, or content updates.
- Safe to upgrade with no impact on current functionality.
v0.3.3
- Simplified and improved first-time setup: activating memory now requires only an email and password, with all manual steps automated by the signup script.
- Updated instructions: the signup script now handles config changes and restarts the gateway automatically—no need for manual environment variable setup.
- Clarified that API credentials are never pasted into chat; all credential management is handled securely by the script.
- Daily usage, privacy, and memory recall instructions remain unchanged.
v0.3.2
spark-memory 0.3.2
- Updated documentation examples to clarify memory recording, removing concrete sensitive data (e.g., property gate codes).
- Privacy section now avoids referencing property-specific sensitive operational data.
- Enhanced focus on role-based business information rather than personal or secret details in memory guidance.
- Minor corrections in recording usage examples for clarity and privacy consistency.
v0.3.1
- Added LICENSE file to the project.
- Added a new script: scripts/spark-signup.sh for guided command-line account signup.
- Updated onboarding instructions: Now guides users to run spark-signup.sh for easy setup, in addition to manual web signup.
- Improved clarity in first-time setup documentation, including shell profile and config guidance.
- Refined memory recording examples for consistency in type labels during conversation.
- No changes to backend or core functionality.
v0.3.0
- Updated backend API endpoint information to use the official `https://zellin.ai/api` domain for Spark.
- Clarified data encryption: all data is encrypted in transit (HTTPS) and at rest.
- Provided improved instructions and API verification details.
- No change to scripts or logic; documentation and privacy details improved.
v0.2.2
- The setup instructions now clearly tell users not to share their API keys in chat; instead, they must set them as environment variables or in a config file for better security.
- The setup sequence has been clarified, emphasizing restarting the gateway and pointing users directly to the environment variable section from the signup page.
- The documentation now explains more precisely that the backend endpoint is the official Zellin Spark API, giving details about hosting, verification, and security.
- References to users sharing API keys over chat have been removed or replaced with best practice instructions.
- Minor documentation updates and clarifications regarding backend infrastructure and security practices.
v0.2.1
**Adds first-time user setup instructions and improves onboarding clarity.**
- Added a dedicated "First Time Setup" section guiding users through account creation, key retrieval, and environment variable setup.
- Clarified that both `SPARK_API_KEY` and `SPARK_ORG_ID` are required, and the skill will prompt for them if missing.
- Provided sample user-facing explanations for what Spark Memory does.
- Minor metadata change: removed `SPARK_API_URL` from required environment variables.
- All previous usage, privacy, and memory handling guidance is retained.
v0.2.0
- Added new insights and dream introspection tools with the scripts/spark-insights.sh script.
- Expanded memory handling guidance, including new categories for recording explicit rules, policy declarations, and correction/frustration cues.
- Improved and clarified privacy, data isolation, and API key permissions documentation.
- Updated requirements to include the SPARK_API_URL environment variable.
- Expanded documentation on nightly "Dream Intelligence," detailing six phases for smarter memory synthesis.
- Provided direct links to privacy policy and source code for transparency.
v0.1.1
- Added explicit mention of Zellin as the provider and clarified Spark's infrastructure ("Powered by Spark by Zellin").
- Introduced required environment variables: `SPARK_API_KEY` and `SPARK_ORG_ID`.
- Added a detailed "Privacy & Data Handling" section specifying what should and should not be recorded, including explicit privacy guidelines and a reference to Zellin's policy.
- Clarified that Spark memory is cloud-hosted and org-scoped, with encryption and data isolation.
- Provided Zellin contact information and privacy policy link for user reference.
v0.1.0
spark-memory 0.1.0
- Initial release introducing persistent, intelligent memory for AI agents via Spark.
- Enables recalling relevant information across sessions for better continuity.
- Includes tools for recording key facts, preferences, decisions, and corrections with importance scoring.
- Introduces daily reflection process to synthesize patterns from accumulated memories.
- Provides scripts for recalling, recording, summarizing, and checking memory status.
元数据
常见问题
Spark Memory 是什么？
Intelligence layer that compounds. Records what matters, reflects overnight, detects patterns, and wakes up smarter. 6-phase dream cycle, proactive morning i... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 204 次。
如何安装 Spark Memory？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install spark-memory」即可一键安装，无需额外配置。
Spark Memory 是免费的吗？
是的，Spark Memory 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Spark Memory 支持哪些平台？
Spark Memory 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 Spark Memory？
由 zellinun（@zellinun）开发并维护，当前版本 v0.3.7。
推荐 Skills