Openclaw Skill Intelligence Ingestion

Intelligence Ingestion Skill

When the user shares a URL, article, tweet, thread, or any piece of external information, execute this pipeline automatically. Do NOT ask for permission — just process it.

Trigger Conditions

USE this skill when: User sends a URL (x.com, github.com, any domain) User pastes article text or tweet content User says "analyze this", "evaluate this", "what do you think about this" User forwards content from Telegram or any messaging surface User shares something and asks about strategic value DO NOT USE when: User is asking a general question unrelated to external content User is asking about internal workspace files User explicitly says they don't want analysis

Step 1: READ — Extract Content

If a URL is provided, read the full content (use read_url_content or browser) If text is pasted, use it directly If X/Twitter link fails to parse (common), search web for the tweet content

Step 2: CLASSIFY — What Is This?

Assign one primary category and up to 2 secondary tags: CategoryDescriptionExamplesinfraInfrastructure / protocols / networkingPilot Protocol, MCP, networking stacksstrategyRouting / cost / architecture decisionsModel routing, multi-account, fallback chainsskillAgent skills / tools / capabilitiesOpenAI Skills, Skill design patternstheoryConceptual frameworks / mental modelsBayes, decision theory, learning loopstutorialHow-to guides / learning materialClaude Code, OpenClaw tutorialsproductTools / apps / servicesLM Studio, new AI models, appscommunityOpenClaw ecosystem / discussionsCommunity posts, feature requeststhreatRisks / security / deprecationAPI changes, breaking updates, security alerts

Step 3: ANALYZE — Strategic Value Assessment

• For each piece of information, evaluate:
• ## Strategic Assessment
• **What is it?** [One sentence]
• **What can it do for us?** [Specific capability/benefit]
• **What can we build with it?** [Concrete output/project]
• **Strategic value:** [🔴 Critical / 🟡 High / 🟢 Medium / ⚪ Low]
• **Competitive edge:** [What advantage over people who don't have this?]
• **Relation to active bottleneck:** [Does it relate to context overflow/token saving?]
• Reference the current engineering bottleneck from MEMORY.md → "Active Engineering Bottleneck" section.

Step 4: MAP — Relate to Existing Architecture

Check against the OpenClaw stack: SOUL.md → PRINCIPLES.md → AGENTS.md (Identity Stack) MEMORY.md (System State + Bottleneck) TOOLS.md (Coprocessors: Codex, Antigravity, LM Studio) Pilot Protocol (Context Separation Layer - P0) Determine which layer this information impacts and note dependencies/synergies with existing components.

Step 5: STORE — Write Obsidian Note

Create a note at: /Volumes/T7 Shield/Obsidian_Vault/20_Intelligence/YYYYMMDD_AuthorOrSource_ShortTitle.md Follow this template exactly: # [Title] **Source:** [Link](URL) **Date:** YYYY-MM-DD **Category:** [Primary Category] / [Secondary Tags] **Strategic Value:** [🔴/🟡/🟢/⚪] + one-line reason **Relation to Active Bottleneck:** [Yes/No + how] --- ## Summary [2-3 paragraphs of core content] ## Key Takeaways [Numbered list of actionable insights] ## Impact on OpenClaw Architecture [How this relates to our stack] ## Action Items [What to do next, if anything] --- **Keke's Note:** [Opinionated analysis in Keke's voice — direct, no-BS, relate to 阳哥's goals]

Step 6: REMEMBER — Update Memory

Always append to today's daily log: ~/.openclaw/workspace/memory/YYYY-MM-DD.md If strategic value is 🔴 Critical: Also update MEMORY.md (Pending Work or Active Bottleneck) If it suggests a new principle: Flag for potential PRINCIPLES.md update If it's a new tool/service: Flag for potential TOOLS.md update

Step 7: RESPOND — Summarize to User

Reply with a concise summary: 📥 已摄取: [Title] 📂 类别: [Category] 🎯 战略价值: [🔴/🟡/🟢/⚪] [One-line reason] 💾 已存档: Obsidian → 20_Intelligence/[filename] 📝 已记录: memory/YYYY-MM-DD.md 🔗 关联: [Which existing component it relates to] ⚡ 建议行动: [Next step, if any]

Edge Cases

Multiple URLs in one message: Process each separately, create separate Obsidian notes Duplicate/similar content: Check if similar note exists, merge or reference instead of duplicating Non-English content: Analyze in original language, write notes in Chinese (matching existing vault style) Paywalled/inaccessible content: Note as "content unavailable" and work with whatever user provided User provides their own analysis: Incorporate their judgment, don't overwrite — they know their system best

Quality Checklist

Before completing, verify: Obsidian note created with correct filename format Daily memory log updated Source URL preserved in note Strategic value assessed against current bottleneck Keke's Note written with genuine analysis (not generic) User received confirmation summary