IntelliClaw

Real-time multi-topic signals pipeline for research, OSINT, job-market monitoring, market-close analysis, and configurable event intelligence.

Built on OpenClaw · Operated by AIML Solutions

---

What It Does

IntelliClaw is an autonomous intelligence pipeline for monitoring configurable topics and event domains. It can be adapted for arXiv/research feeds, job-search signals, geopolitics, markets, infrastructure, incident response, and historical timelines. It normalizes multilingual content, cross-checks claims for contradictions, scores signals by risk level, and dispatches structured intelligence updates to a live ledger.

In the MultiClaw OS architecture, IntelliClaw is the intelligence layer: public-source signal collection and triage that can feed agent workflows, research briefs, hiring pipelines, market context, or client-defined watchlists.

Who This Is For

• Analysts and engineers who need configurable research monitoring without a heavyweight platform
• Recruiters or clients evaluating Python/bash automation, source hygiene, and signal-triage design
• Teams building public/private intelligence workflows where credentials, client notes, and sensitive claims must stay separated

Portfolio Proof

Area	Evidence
Source configuration	RSS source list and source-class model
Pipeline orchestration	shell orchestrator across harvesting, normalization, cross-checking, scoring, and dispatch
Public/private boundary	README and docs require sanitized examples only
Example artifacts	examples/signal-cycle.input.json, examples/signal-cycle.output.example.json
Next proof target	fixture-based complete run with tests

Example Lanes

• arXiv and technical paper monitoring
• job-search and target-company intelligence
• market close summaries and macro/derivatives signal collection
• OSINT and incident monitoring
• infrastructure and internet availability signals
• custom client or research-topic watchlists

Public examples should stay sanitized. Credentials, private job-search notes, client details, and account-specific market data do not belong in this repository.

Commercial Use Cases

• Research monitoring for markets, technical papers, jobs, incidents, or client-defined watchlists
• Sanitized intelligence summaries with source and confidence metadata
• Job-search and target-company monitoring as a private workflow
• OSINT-style source normalization and contradiction triage with explicit caveats

Implementation Status

Component	Status	Notes
Feed harvesting	Active	RSS harvesting and source configuration are implemented under skills/intelliclaw-feed-harvester/ and operations/IntelliClaw/config/
Normalization	Active	Language/entity normalization skill structure is present
Cross-checking	Scaffold	Current contradiction checks are structured but should be expanded with entity/time/source comparison logic
Risk scoring	Active/scaffold	Keyword and confidence scoring exists; calibration and evaluation examples are next steps
Telegraph/minutes output	Active	Dispatch and cycle-summary skills are included
Orchestration	Active	A shell orchestrator coordinates the lane sequence for scheduled or manual runs

IntelliClaw is best presented as a configurable signal-triage and research-monitoring pipeline. The next quality step is adding fixtures, CI, and sample outputs that demonstrate a complete run end-to-end.

Pipeline Shape

RSS Feeds (7 sources)
       │
       ▼
┌─────────────────┐
│  feed-harvester │  pulls & parses RSS → raw-claims.json
└────────┬────────┘
         │
         ▼
┌──────────────────────┐
│  persian-normalizer  │  entity normalization, FA detection
└────────┬─────────────┘
         │
         ▼
┌───────────────────┐
│  claim-crosscheck │  contradiction detection → crosscheck-report.json
└────────┬──────────┘
         │
         ▼
┌──────────────┐
│  risk-scorer │  confidence × keyword boost → scored-claims.json
└──────┬───────┘
       │
       ▼
┌───────────────────┐
│ telegraph-writer  │  dispatches → telegraph-ledger.md
└──────┬────────────┘
       │
       ▼
┌────────────────┐
│ minutes-scribe │  cycle summary → running-minutes.md
└────────────────┘

Cycle time: 10 minutes (cron) · Claims per cycle: configurable · Sources: configurable

---

Sources

Label	Class	Coverage
Reuters-World	international	Wire service
AP-World	international	Wire service
BBC-World	international	Global coverage
Al-Jazeera	international	Regional/global analysis
Financial-Times-Markets	markets	Markets and macro coverage
NetBlocks-Global	sensor	Infrastructure/internet signals
Event-Topic-Feed	configurable	User-selected topic stream

---

Quick Start

Requirements

• Python 3.10+
• jq
• bash
• curl

Install

git clone https://github.com/AIML-Solutions/intelliclaw.git
cd intelliclaw

Run a single cycle

bash skills/intelliclaw-orchestrator/scripts/run_intelliclaw_orchestrator.sh .

Run the offline fixture demo

The fixture demo requires no network access and writes deterministic public-safe outputs under examples/demo-output/:

python3 scripts/run_fixture_demo.py
python3 -m unittest discover -s tests

It demonstrates raw claim loading, normalization, contradiction flagging, risk scoring, and a minutes artifact using sanitized example inputs.

Run every 10 minutes (cron)

crontab -e

Add:

*/10 * * * * cd /path/to/intelliclaw && bash skills/intelliclaw-orchestrator/scripts/run_intelliclaw_orchestrator.sh . >> operations/IntelliClaw/live/cycle.log 2>&1

Check dependencies

bash operations/IntelliClaw/scripts/check_dependencies.sh

---

Configuration

Edit operations/IntelliClaw/config/rss_sources.txt to add or remove sources:

# label|class|url
Reuters-World|international|https://...

Supported classes: international, state, opposition, sensor, ugc

Each class maps to a base confidence score. See docs/CONFIGURATION.md.

---

Output Files

File	Description
live/raw-claims.json	Raw harvested claims
live/normalized-claims.json	Normalized and language-tagged claims
live/crosscheck-report.json	Contradiction analysis
live/scored-claims.json	Risk-scored claims
live/intelliclaw-telegraph-ledger.md	Live intelligence dispatches
live/intelliclaw-running-minutes.md	Cycle-by-cycle summary log
live/cycle.log	Cron execution log

Public-safe example artifacts are available under examples/.

Committed fixture-demo outputs are available under examples/demo-output/.

---

Roadmap

• RSS harvest pipeline (7 sources)
• Multilingual normalization
• Claim cross-check
• Risk scoring
• Telegraph ledger dispatch
• 10-min autonomous cron cycle
• SignalCockpit integration (browser auth)
• Contradiction persistence across cycles
• Cross-cycle deduplication
• One-pager prose summary
• Optional translation toggle for multilingual streams
• Web dashboard
• Public API

---

Project Structure

intelliclaw/
├── skills/
│   ├── intelliclaw-feed-harvester/
│   ├── intelliclaw-persian-normalizer/
│   ├── intelliclaw-claim-crosscheck/
│   ├── intelliclaw-risk-scorer/
│   ├── intelliclaw-telegraph-writer/
│   ├── intelliclaw-minutes-scribe/
│   └── intelliclaw-orchestrator/
├── operations/
│   └── IntelliClaw/
│       ├── config/
│       │   └── rss_sources.txt
│       ├── scripts/
│       │   └── check_dependencies.sh
│       └── live/          ← gitignored outputs
└── docs/
    ├── ARCHITECTURE.md
    ├── PIPELINE.md
    ├── SKILLS.md
    ├── CONFIGURATION.md
    ├── CONTRIBUTING.md
    └── INTELLIGENCE-LAYER.md

---

License

MIT © 2026 AIML Solutions

Live

The pipeline runs every six hours on GitHub Actions (pure RSS + jq, no API keys, no model spend) and publishes a bounded public snapshot:

• public/latest.json — cycle counts, source breakdown, top high-risk signals
• public/onepager.md — the one-page digest

The snapshot is rendered live on aiml-solutions.com. Run a cycle locally with bash skills/intelliclaw-orchestrator/scripts/run_intelliclaw_orchestrator.sh "$PWD" && python3 scripts/export_public.py.