← 返回 Skills 市场
284
总下载
0
收藏
4
当前安装
2
版本数
在 OpenClaw 中安装
/install recall
功能描述
Teaches agents to check before they guess. Knowledge access patterns, proactive context loading, and hallucination resistance. Install before taking the Know...
安全使用建议
This skill is coherent and instruction-only: it tells an agent to load workspace context and check installed skills before answering. Before installing, verify that your workspace contains only files you want agents to read (move or protect any .env, .secrets, or credentials), and confirm you trust other installed skills because this skill instructs the agent to scan skills/ and read their SKILL.md. There's a minor manifest version mismatch (manifest.json lists 1.0.0 while registry metadata is 1.1.0) — likely harmless but worth noting. If you limit agent filesystem access in your environment, ensure the agent can still read the intended agent-readable docs (AGENTS.md, TOOLS.md, memory/) or the skill's benefits will be reduced.
功能分析
Type: OpenClaw Skill
Name: recall
Version: 1.1.0
The 'recall' skill bundle consists entirely of instructional markdown files designed to improve an AI agent's accuracy by encouraging it to check local workspace files and tools before answering. It includes explicit safety guidelines in SKILL.md and references/session-checklist.md that instruct the agent NOT to proactively read sensitive files like .env or .secrets unless specifically directed by the user. There is no executable code, no evidence of data exfiltration, and the behavioral instructions are transparently aligned with the stated goal of reducing hallucinations.
能力评估
Purpose & Capability
Name/description (teach agents to check local context and avoid hallucination) align with the SKILL.md instructions, which focus on reading workspace files, scanning installed skills, and loading memory. It does not request unrelated credentials, binaries, or external services.
Instruction Scope
The runtime instructions direct the agent to read specific workspace files and directories (AGENTS.md, TOOLS.md, memory/, skills/, tasks/, etc.) — this is coherent with the skill's goal. The SKILL.md explicitly warns not to proactively read .env or secret stores. Note: the security of this behavior depends on the agent enforcing that policy; proactive file reads could expose sensitive data if workspace files contain secrets or are misnamed.
Install Mechanism
No install spec and no code files that get written or executed — instruction-only, minimal disk/system footprint.
Credentials
The skill declares no required environment variables, credentials, or config paths. The SKILL.md references workspace files but explicitly disallows proactive reading of obvious secret locations (e.g., .env). Requested access is proportional to the stated purpose.
Persistence & Privilege
always is false and the skill is user-invocable; it does not request permanent inclusion or modify other skills. Autonomous invocation is allowed (platform default) but not combined with any broad credential access or install behavior that would increase risk.
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 recall - 安装完成后，直接呼叫该 Skill 的名称或使用
/recall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
v1.1.0
Recall v1.1.0
- Added explicit version and homepage fields to metadata.
- Clarified context loading: only read files intended for agent consumption; avoid proactive access to secrets, credentials, or `.env` files.
- Updated examples and guidelines to reinforce access restrictions to sensitive files, aligning with privacy and security best practices.
- Minor content improvements for clarity and precision in several disciplines.
v1.0.0
- Initial release of the "recall" skill, focused on enhancing agents’ knowledge access patterns and reducing reliance on training data.
- Introduces five disciplines: Check Before You Guess, Proactive Context Loading, Tool-First Problem Solving, Hallucination Resistance, and Multi-Source Synthesis.
- Designed to improve accuracy, confidence signaling, and proactive preparation by referencing available files, skills, and tools before responding.
- Provides detailed instructions and real examples for each discipline, promoting best practices and reducing hallucination.
- Aimed at agents preparing for the Knowledge-Driven Agent certification.
元数据
常见问题
Recall 是什么？
Teaches agents to check before they guess. Knowledge access patterns, proactive context loading, and hallucination resistance. Install before taking the Know... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 284 次。
如何安装 Recall？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install recall」即可一键安装，无需额外配置。
Recall 是免费的吗？
是的，Recall 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Recall 支持哪些平台？
Recall 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 Recall？
由 crewhaus（@crewhaus）开发并维护，当前版本 v1.1.0。
推荐 Skills