← 返回 Skills 市场
ianderrington

Agent Knowledge Capture

作者 ianderrington · GitHub ↗ · v1.0.0
cross-platform ⚠ suspicious
1824
总下载
1
收藏
10
当前安装
1
版本数
在 OpenClaw 中安装
/install agent-knowledge
功能描述
Unified knowledge capture and retrieval for URLs, video/article/paper extracts, social posts, and agent research outputs. Use when saving anything worth re-f...
使用说明 (SKILL.md)

Knowledge Skill

File-based knowledge organization. Capture fast, search later, clean up automatically.

Installation

clawhub install knowledge

This installs scripts/know — add to PATH or use full path:

~/.openclaw/skills/knowledge/scripts/know

Storage Location

Default: ~/.soulshare/agent/knowledge/

Configurable via ~/.config/know/config or env KNOWLEDGE_DIR.

knowledge/
├── INDEX.md      # Auto-maintained browsable index
├── urls/         # Bookmarked URLs
├── extracts/     # Video/article/paper summaries
├── posts/        # Social content (tweets, threads)
└── research/     # Agent-generated research

Adding Content

know add url \x3Curl> --title "..." --tags "a,b" [--summary "..."]
know add video \x3Curl> --title "..." --tags "a,b" [--summary "..."]
know add extract --source \x3Curl> --type video|article|paper --title "..." --tags "a,b"
know add post --source \x3Curl> --title "..." --tags "a,b"
know add research --title "..." --tags "a,b" [--summary "..."]

Each add writes a markdown file with YAML frontmatter and updates INDEX.md.

Searching

know search "query"           # Grep across all entries
know recent --limit 10        # Recent entries
know list --tags security     # Filter by tag

Cleanup & Maintenance

know tidy                     # Audit: find issues
know tidy --fix               # Auto-fix: normalize tags, move misplaced files, remove empty
know validate                 # Check frontmatter schema
know reindex                  # Rebuild INDEX.md
know config                   # Show active config paths

Recommended: Run know tidy --fix in heartbeats or nightly cron.

Data Model (frontmatter)

---
type: url|extract|post|research
title: "Entry title"
source_url: "https://..."
source_kind: url|video|article|paper|post|research
tags: ["tag1", "tag2"]
added: "2026-02-26"
added_by: "agent-name"
summary: "One-line summary"
---

QMD Integration

Plain markdown → QMD-indexable.

qmd collection add ~/.soulshare/agent/knowledge --name knowledge
qmd search "query" --collection knowledge

Discipline

  • If it's useful later → know add immediately
  • Don't leave knowledge only in memory/YYYY-MM-DD.md
  • Every entry needs tags + summary
  • Let know tidy --fix handle normalization
安全使用建议
This skill appears to document a local CLI for saving and maintaining notes, but the package contains only documentation and no executable or install spec. Before installing or enabling automation: 1) Confirm the origin — where does 'clawhub install knowledge' download from and is that source trustworthy (there's no homepage or source listed). 2) Verify the actual 'know' binary and inspect its code before running it, since the SKILL.md expects it to modify files under ~/.soulshare/agent/knowledge/ and run tidy --fix (which can remove or move files). 3) Back up your existing ~/.soulshare (and any target directories) before running auto-fix/cron tasks. 4) If you don't want automated modifications, avoid scheduling tidy --fix in heartbeats/cron and run maintenance manually after inspection. 5) Ask the publisher for the missing install manifest or source code; lack of code + install instructions is the main reason this is flagged as suspicious.
功能分析
Type: OpenClaw Skill Name: agent-knowledge Version: 1.0.0 The `SKILL.md` describes a knowledge management skill that relies on an external script named `know` (located at `scripts/know`), which is not provided for analysis. The documentation mentions `know search` using `grep` for querying, which is a common pattern that can lead to shell injection vulnerabilities if the `know` script does not properly sanitize user input before executing `grep`. Without the `know` script, the full security posture cannot be assessed, but the described functionality presents a potential risk for command injection.
能力评估
Purpose & Capability
Name/description (knowledge capture and retrieval) align with the instructions (commands to add/search/tidy knowledge entries). However the SKILL.md references an install command (clawhub install knowledge) and an on-disk script (~/.openclaw/skills/knowledge/scripts/know) even though the registry entry contains no install spec or code files. That gap is an incoherence to investigate.
Instruction Scope
Instructions stay within the stated domain: adding entries, searching, and maintaining a local knowledge directory. They explicitly write files under ~/.soulshare/agent/knowledge/ and read an optional config at ~/.config/know/config or env KNOWLEDGE_DIR, and recommend running tidy --fix periodically (which can move/remove files). Those file operations are expected for this purpose but are destructive by design (auto-fix/remove), so they should be reviewed before enabling automated runs.
Install Mechanism
The SKILL.md claims an installation step and a bundled script path, yet the skill package contains no install spec and no code files. Because there is no provided install mechanism or code, it's unclear where the 'know' binary comes from, what 'clawhub install knowledge' would fetch, or whether external downloads are required — this is an important inconsistency and a potential supply-chain risk if you try to follow the install instructions.
Credentials
No credentials or sensitive environment variables are requested. The only environment/config references are an optional KNOWLEDGE_DIR and a config file path, which are proportional to a local knowledge-management tool.
Persistence & Privilege
The skill does not request always-on privilege (always: false). It recommends running periodic maintenance (heartbeats/cron) which would allow automated modification/deletion of files; if enabled, that increases risk. The skill does not request access to other skills' configs or system-wide credentials.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-knowledge
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-knowledge 触发
  4. 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: file-based knowledge capture CLI for AI agents. Includes know add/search/tidy commands. Storage: ~/.soulshare/agent/knowledge/
元数据
Slug agent-knowledge
版本 1.0.0
许可证
累计安装 11
当前安装数 10
历史版本数 1
常见问题

Agent Knowledge Capture 是什么?

Unified knowledge capture and retrieval for URLs, video/article/paper extracts, social posts, and agent research outputs. Use when saving anything worth re-f... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 1824 次。

如何安装 Agent Knowledge Capture?

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

Agent Knowledge Capture 是免费的吗?

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

Agent Knowledge Capture 支持哪些平台?

Agent Knowledge Capture 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 Agent Knowledge Capture?

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

💬 留言讨论