菜单
Build and maintain structured, persistent knowledge graphs across sessions. Knowledge-ops transforms ephemeral session findings into reusable intelligence — connecting entities, tracking confidence over time, and enabling recall across engagements.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/brucesongs/kali-claw --skill knowledge-ops命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | knowledge-ops |
| description | Build and maintain structured, persistent knowledge graphs across sessions. Knowledge-ops transforms ephemeral session findings into reusable intelligence — connecting entities, tracking confidence over time, and enabling recall across engagements. |
| origin | openclaw |
| version | 0.2.0.2 |
| compatibility | ["openclaw","claude-code","cursor","windsurf"] |
| allowed-tools | ["Bash","Read","Write","Edit","WebSearch","WebFetch","Agent"] |
| defense_triple_required | false |
| metadata | {"domain":"knowledge","tool_count":0,"guide_count":5,"last_reviewed":"2026-07-26"} |
Knowledge-ops transforms ephemeral session findings into reusable intelligence — connecting entities, tracking confidence over time, and enabling recall across engagements.
Domain: knowledge
| Attribute | Value |
|---|---|
| Domain | Knowledge Operations |
| Skill ID | knowledge-ops |
| Version | 1.0.0 |
| Hacker Laws | Law 3 (Intelligence Over Force), Law 8 (Learn from Every Operation), Law 9 (Systematic Over Random) |
| Related Skills | codebase-onboarding, deep-research, social-intelligence |
Build and maintain structured, persistent knowledge graphs across sessions. Knowledge-ops transforms ephemeral session findings into reusable intelligence — connecting entities, tracking confidence over time, and enabling recall across engagements.
Without knowledge-ops, every session starts from zero. With it, each session builds on the last.
The atomic element of the knowledge graph. Each unit contains:
{
"id": "KU-2026-05-001",
"type": "finding | entity | relationship | pattern | hypothesis",
"created": "2026-05-11",
"updated": "2026-05-11",
"confidence": 78,
"tags": ["auth", "JWT",
A network of linked knowledge units. Relationships between units reveal:
A collection of knowledge graphs organized by domain (target, technology, vulnerability type). Stored in the memory/ directory as markdown files with structured frontmatter.
| Type | Description | Example |
|---|---|---|
entity | A person, system, domain, IP, org | target.com, admin@target.com, 192.168.1.1 |
finding | A discovered fact about a target | SQL injection in search endpoint |
relationship | Link between two entities | User X administers System Y |
pattern | Recurring vulnerability or behavior | This org consistently uses outdated JWT libs |
hypothesis | Unconfirmed, needs validation | Admin portal may be accessible via VPN bypass |
intelligence | Aggregated, analyzed insight | Target org has poor patch management practices |
Extract knowledge units from session outputs:
Assign type, tags, confidence, and links to each unit. Establish relationships between related units.
Write structured units to the appropriate memory file. Follow the naming convention: memory/YYYY-MM-DD-[topic].md.
At session start, query the knowledge base for relevant units. Use tags, entity names, and linked IDs to surface prior context.
Update confidence scores as findings are confirmed or refuted. Mark expired intelligence. Remove stale hypotheses that were disproven.
Confidence scores represent certainty about a knowledge unit:
| Score | Meaning | Action |
|---|---|---|
| 0–25 | Speculation — no evidence | Label as hypothesis |
| 26–50 | Unconfirmed — single source | Needs validation |
| 51–75 | Probable — multiple consistent signals | Use with caution |
| 76–90 | High confidence — independently confirmed | Act on it |
| 91–100 | Verified — demonstrated/exploited | Treat as fact |
Confidence is not static. Update it when:
---
id: KU-[YYYY]-[MM]-[NNN]
type: [finding|entity|relationship|pattern|hypothesis|intelligence]
target: [target name or scope]
confidence: [0-100]
created: [YYYY-MM-DD]
updated: [YYYY-MM-DD]
tags: [tag1, tag2, tag3]
linked: [KU-ID-1, KU-ID-2]
source: [skill that produced this]
expires: [YYYY-MM-DD or null]
---
## Detail
[Full context, evidence, commands used, observed behavior]
## Connections
- **Confirms**: [what this validates]
- **Contradicts**: [what this disproves]
- **Leads to**: [next investigation or action]
## Confidence History
| Date | Score | Reason |
|------|-------|--------|
| [date] | [score] | [why this score] |
index=kb sourcetype=document_access | stats dc(doc_id) by user_id | where dc > 100