Intelligence Ingestion

Intelligence Ingestion Skill

You are feeding something smarter than you. This Skill ensures you know exactly what it ate and what it became. MCP Compatible: This Skill ships with a manifest.json that follows the MCP capability declaration spec, making it discoverable and invokable by any MCP-compatible Agent workflow.

Design Philosophy

"giving my private data/keys to 400K lines of vibe coded monster is not very appealing at all" — Andrej Karpathy This Skill embraces a trust-first design philosophy: No Self-Modifying Agents. Auto-generated Skills are isolated as drafts. You, the human operator, must review and approve them before they activate. Skills are the new Config. No need to modify configuration files or monster if-else scripts. A single Skill file defines a capability. Build. For. Agents. Built with an MCP manifest, structured outputs, and CLI installation so that any Agent in the ecosystem can find, understand, and use it.

Permissions & Privacy

This Skill writes files to your local filesystem and may access sensitive browser state. Full transparency on what it touches: PermissionWhatWhyFile Write{obsidian_vault}/{intelligence_folder}/Creates structured intelligence notesFile Write{workspace}/STRATEGIC_LANDSCAPE.mdUpdates capability map when critical info is ingestedFile Write{workspace}/skills/_drafts/Creates auto-synthesized Skill drafts (isolated, never auto-loaded)File Write{workspace}/memory/Appends to daily memory logsFile CreateSTRATEGIC_LANDSCAPE.mdAuto-creates from template if missing on first runFile Read{workspace}/skills/Reads existing Skills for gap analysis during synthesisNetworkUser-provided URLsFetches content via HTTP; may fall back to search if primary fetch failsCredentialxurl X API auth (Optional)Used for authenticated X/Twitter API v2 access via the xurl Skill Privacy Note: The "never return empty-handed" fallback policy means the Skill may make multiple external network requests per ingestion (direct fetch → xurl → web search). All fetched content is stored locally in your Obsidian vault; no data is transmitted to third-party servers beyond the original fetch.

Prerequisites

ToolPurposeRequired?ObsidianKnowledge storage (notes land here)✅ RequiredOpenClawSkill host + Agent orchestration✅ Required

Configuration

Edit config.json after installation: { "obsidian_vault_path": "/path/to/your/Obsidian_Vault", "intelligence_folder": "20_Intelligence", "landscape_path": "STRATEGIC_LANDSCAPE.md", "output_dir": "/path/to/your/output" } FieldDescriptionDefaultobsidian_vault_pathAbsolute path to Obsidian Vault rootNone (Required)intelligence_folderSubfolder for intelligence notes20_Intelligencelandscape_pathStrategic Landscape file path (relative to workspace)STRATEGIC_LANDSCAPE.mdoutput_dirDirectory for non-knowledge outputsNone (Optional) Storage Strategy: Obsidian = Knowledge Inputs (Analysis/Notes), Output = Content Generation (Posts/Scripts)

First-time Setup

If STRATEGIC_LANDSCAPE.md does not exist, the Skill will automatically generate an empty template. If the Obsidian Vault path does not exist or config.json is missing, the Skill will error and provide remediation steps.

Triggers

Auto-triggers on: User sharing a URL (x.com, github.com, arxiv.org, etc.) User pasting an article/tweet block User prompts like "analyze this", "evaluate this", "what do you think about this" User forwarding content from Telegram or any chat interface Does not trigger on: General questions unrelated to external knowledge Explicit commands to simply summarize E-commerce, music, or entertainment links

Content Extraction Strategy (Critical)

Different sources require different extraction methods. The Agent must run down this priority chain:

Standard Web Pages

1. read_url_content(url) → If successful, use it 2. Fallback → Direct user to paste content

X/Twitter (Aggressive Anti-Scraping)

X strictly prevents unauthenticated scraping. Use this degradation chain: 1. xurl Skill (Recommended) → Native OpenClaw skill using X API v2 - Most reliable method - Requires xurl authentication 2. Fallback → Web search the tweet text (often indexed by search engines) 3. Fallback → Ask user to paste the raw text Principle: Never return empty-handed. Even if extraction fails, inform the user exactly which step failed, why, and how to fix it.

GitHub

1. read_url_content(url) → Usually accessible 2. If repo root → Target README.md 3. If issue/PR → Target title, description, and primary comments

Paywalls / Login Walls

1. Mark as unreachable and analyze based solely on user-provided summary

Step 1: READ — Extract Content

Execute extraction according to the Strategy chain. Log the successful extraction method.

Step 2: CLASSIFY — Categorize

Assign 1 primary category + up to 2 tags: CategoryDescriptionExampleinfraInfrastructure / protocols / networkingMCP, Pilot ProtocolstrategyArchitectural decisions / routing / cost opModel routing, multi-accountskillAgent Skills / tools / capabilitiesSkill patterns, MCP interfacesbusinessBusiness models / market signalsSaaS frameworks, pricingtheoryConceptual frameworks / mental modelsBayes, decision theorytutorialLearning material / guidesClaude Code, Agent LoopsproductNew tool/service releasesLM Studio, model dropsthreatRisk / security / deprecationsAPI changes, vulnerabilities

Step 3: ANALYZE — Strategic Valuation

• Formulate the following breakdown:
• ## Strategic Assessment
• **What is it?** [One-sentence summary]
• **Value Proposition:** [Specific capability/benefit]
• **Actionability:** [Specific outputs/projects it enables]
• **Strategic Value:** [🔴 Critical / 🟡 High / 🟢 Medium / ⚪ Low]
• **Competitive Advantage:** [What is the cost of NOT knowing this?]
• **Capability Boundary Shift:** [What can the Agent do now that it couldn't do before?]

Step 4: MAP — Correlate with Architecture

Read {landscape_path} (Strategic Landscape). Answer: What architectural layer does this impact? What existing components does this depend on? Does the Landscape map need to be updated? (If Landscape file does not exist, skip and prompt user in Step 8 to initialize.)

Step 5: STORE — Mint Obsidian Note

Target Path: {obsidian_vault_path}/{intelligence_folder}/YYYYMMDD_Source_Title.md Template: # [Title] **Source:** [Link](URL) **Date:** YYYY-MM-DD **Category:** [Primary] / [Tags] **Strategic Value:** [🔴/🟡/🟢/⚪] + reason **Extraction Method:** [read_url / browser / search / user_paste] --- ## Abstract [2-3 paragraph core summary] ## Key Takeaways [Numbered list] ## Architectural Impact [Relationship with current systems] ## Capability Boundary Shift [What the Agent can do now vs before. If none, write "None"] ## Action Items [Next steps. If none, write "N/A"] --- **Analytical Notes:** [Direct, sharp, opinionated analysis]

Step 6: SYNTHESIZE — Auto-Skill Generation (Evolution Step)

• Trigger: Content describes an actionable tool, API, protocol, or technique that the Agent currently lacks.
• Do NOT trigger if: Content is purely theoretical, an opinion piece, a capability the Agent already possesses, or ranked ⚪ Low purely regarding strategic value.
• Execution:
• Gap Analysis: Compare current skills/ directory and Strategic Landscape to confirm this is a missing capability.
• Draft Minting: Create a new Skill folder in {workspace}/skills/_drafts/:
• skills/_drafts/[skill-name]/
• ├── SKILL.md # Generated behavior definition
• └── README.md # Brief description, source, dependencies
• SKILL.md Template:
• ---
• name: [Skill Name]
• version: 0.1.0-draft
• description: >
• [Generated description based on ingestion]
• Auto-synthesized by Intelligence Ingestion from: [Source URL]
• ---
• # [Skill Name]
• > 🧬 This Skill was auto-generated by Intelligence Ingestion and requires human review.
• > Source: [URL]
• > Generated: YYYY-MM-DD
• ## Prerequisites
• [Dependencies extracted from content: API Keys, CLI tools, services]
• ## Trigger Conditions
• [When should this trigger based on capability description]
• ## Execution Flow
• [Step-by-step logic pulled from ingested docs]
• ## Input/Output
• **Input:** [Expected input]
• **Output:** [Expected output]
• Security Isolation: Drafts remain in _drafts/ and are ignored by the Agent runtime. The operator must review the code and manually move it to the skills/ directory to activate it.

Step 7: REMEMBER — Update Core Memory

Mandatory: Append to daily log {workspace}/memory/YYYY-MM-DD.md If 🔴 Critical: Sync update to {landscape_path} If Tool related: Flag TOOLS.md for review If Skill Draft Synthesized: Log the draft path and flag it as entirely pending human review.

Step 8: RESPOND — Feedback to Operator

Respond exactly in this layout: 📥 Ingested: [Title] 📂 Class: [Category] 🎯 Value: [🔴/🟡/🟢/⚪] [One sentence] 🔄 Cap Shift: [Delta explanation / None] 🧬 Skill Draft: [Generated → skills/_drafts/[name]/ | N/A] 💾 Archived: Obsidian → {intelligence_folder}/[Filename] 🗺️ Landscape: [Updated / Unchanged / Missing] ⚡ Action: [Review draft / Review landscape / None]

Companion: Strategic Landscape

Intelligence Ingestion pairs directly with your Strategic Landscape: Intelligence Ingestion = The "Information Input" Strategic Landscape = The "Capability Map" Every time 🔴 Critical information is ingested, the Landscape updates automatically, providing a real-time view of what the Agent knows, what it can do, and what the blind spots are.

Input

Analyze this: https://x.com/kaboraAI/status/1234567890

Execution

Step 1: READ → xurl Skill invoked → Success → Content: "MCP is the new TCP for agents. 97M downloads..." Step 2: CLASSIFY → Primary: infra → Tags: skill, strategy Step 3: ANALYZE → Value: 🔴 Critical → Cap Shift: Skills must adopt MCP interfaces immediately. Step 4: MAP → Impact Layer: Layer 2 (Communication) + Layer 4 (Orchestration) → Update Landscape: Yes Step 5: STORE → Obsidian Node: 20_Intelligence/20260225_kaboraAI_MCP_Agent_TCP.md Step 6: SYNTHESIZE → Gap: Agent currently lacks automatic MCP schema generation → Draft: skills/_drafts/mcp-schema-generator/SKILL.md → Logic: Reads existing Skills and auto-generates MCP JSON Step 7: REMEMBER → Updated memory/2026-02-25.md → Updated STRATEGIC_LANDSCAPE.md → Logged Draft: mcp-schema-generator pending review Step 8: RESPOND

Output

📥 Ingested: MCP is the new TCP for agents 📂 Class: infra / skill, strategy 🎯 Value: 🔴 Critical — MCP passing 97M downloads signals 24-month standardization window 🔄 Cap Shift: Skills require MCP interfaces to be discovered by external swarms 🧬 Skill Draft: Generated → skills/_drafts/mcp-schema-generator/ → Function: Auto-compiles OpenClaw Skills to MCP schemas → Review: mv skills/_drafts/mcp-schema-generator skills/mcp-schema-generator 💾 Archived: Obsidian → 20_Intelligence/20260225_kaboraAI_MCP_Agent_TCP.md 🗺️ Landscape: Updated (Communication + Orchestration layers) ⚡ Action: Review schema generator draft; plan fleet-wide audit

Edge Cases

ConditionHandlerMultiple URLsProcess separately; create independent notesRedundant infoMap to existing note; append or referenceNon-EnglishAnalyze raw content; write notes in user's native tongueUnreachable URLMark as failed; analyze whatever context user providedUser provided analysisIntegrate with assessment; do not overrideconfig.json missingHalt; stream remediation stepsLandscape missingSkip Step 4; prompt initialization in Step 8All X/Twitter falls failAsk user to paste textPure theory/opinionSkip SYNTHESIZE; mark N/ACapability existsSkip SYNTHESIZE; mark existing_drafts/ missingAuto-create {workspace}/skills/_drafts/

Quality Checklist

Before responding, verify (using ls or checks): Obsidian note minted as YYYYMMDD_Source_Title.md Daily memory log appended Source URL retained in metadata Strategic Value scored Capability Shift assessed Skill draft synthesized (if applicable, verify _drafts/ path exists) Strategic Landscape updated (if applicable) Operator received standard Step 8 output Explicit fallback reasoning provided if extraction failed