Research focus evolution agent. Extracts user intent from natural language feedback and proposes structured updates to FOCUS.md. Processes accumulated evolution signals and user-expressed interests to refine the research profile. Examples: <example> Context: User read a research report and wants to narrow their focus. user: "I'm interested in the regulatory angle, less so in the technical implementation details" assistant: "I'll use the focus-evolver agent to propose FOCUS.md updates." </example> <example> Context: Accumulated signals suggest new angles after incremental updates. user: "Review the evolution signals and suggest focus changes" assistant: "I'll use the focus-evolver agent to process signals and propose changes." </example>

Deep analysis agent for domain intelligence. Applies source-specific prompts to extract structured insights from raw collected items. Two-stage: quick screen → deep analysis. Produces significance-scored, tagged insight records. Supports ten source types: GitHub repos, Product Hunt launches, RSS articles, official changelogs, notable figures, company news, academic papers, YouTube videos, community discussions, and general web articles. Uses LENS.md context for personalized relevance calibration when available. Examples: <example> Context: Raw GitHub items from source-scanner need deep analysis. user: "Analyze these 15 GitHub items for the configured domains" assistant: "I'll use the insight-analyzer agent to perform deep analysis on the GitHub items." </example> <example> Context: RSS articles need structured insight extraction. user: "Analyze these RSS articles and extract insights" assistant: "I'll use the insight-analyzer agent to analyze the RSS articles." </example> <example> Context: Figure-sourced items need analysis with LENS context. user: "Analyze these figure items about Geoffrey Hinton and Chris Lattner" assistant: "I'll use the insight-analyzer agent to analyze the figure items with LENS context." </example>

Multi-source web collection agent for topic-focused deep research. Searches across search engines, GitHub, academic sources, YouTube, community forums, industry media, official sites, and institutional sources. Returns structured data for filtering and analysis — no judgment, no scoring. Supports broad (initial) and targeted (incremental) scan modes. Examples: <example> Context: First-time deep research on a topic needs broad collection. user: "Collect everything about OpenCLaw from all internet sources" assistant: "I'll use the research-scanner agent to do a broad multi-source collection." </example> <example> Context: Incremental update with evolved focus needs targeted collection. user: "Collect new items about OpenCLaw focusing on regulatory compliance angle" assistant: "I'll use the research-scanner agent in targeted mode with weighted angles." </example>

Report-oriented synthesis agent for topic research. Reads analyzed findings and produces comprehensive research reports with entity extraction, opinion spectrum analysis, timeline construction, and information gap identification. Two modes: comprehensive (initial research) and incremental (update). Examples: <example> Context: First deep research completed, needs comprehensive report. user: "Synthesize 35 findings about OpenCLaw into a comprehensive research report" assistant: "I'll use the research-synthesizer agent to produce a structured report." </example> <example> Context: Incremental update with 8 new findings needs connection to previous research. user: "Generate an incremental report connecting 8 new findings to previous OpenCLaw research" assistant: "I'll use the research-synthesizer agent in incremental mode." </example>

Parallel web collection agent for domain intelligence. Fetches raw items from GitHub trending, RSS feeds, official changelogs, notable figures, company news, and Product Hunt launches. Returns structured data for filtering and analysis — no judgment, no scoring. Examples: <example> Context: Scheduled scan needs fresh items from all configured sources. user: "Collect items from GitHub, RSS feeds, official changelogs, figures, companies, and Product Hunt" assistant: "I'll use the source-scanner agent to collect from all configured sources." </example>

Cross-insight pattern detection and synthesis agent for domain intelligence. Reads multiple insights and produces trend analysis, convergence detection, and collective wisdom. Two modes: general synthesis (for digests) and query-directed (for answering specific questions). Examples: <example> Context: Daily digest needs synthesis of recent insights. user: "Synthesize trends from today's 12 insights across iOS development and AI/ML domains" assistant: "I'll use the trend-synthesizer agent to detect patterns and generate the synthesis." </example> <example> Context: User asks a question about collected intelligence. user: "What's the trend around on-device AI based on recent insights?" assistant: "I'll use the trend-synthesizer agent to synthesize an answer from accumulated insights." </example>

Use when the user says 'digest', 'generate report', 'weekly summary', or when invoked by cron. Generates a daily or weekly digest from accumulated insights by dispatching trend-synthesizer for pattern detection and synthesis.

Use when the user says 'intel', 'briefing', 'what's new', 'intel status', 'intel setup', 'intel config', or asks a question about collected domain insights. Single human-facing entry point for domain intelligence: status, briefings, Q&A, configuration, and exploration.

Use when the user says 'research', 'deep research', 'research topic', or wants comprehensive internet-wide investigation of a specific topic. Supports full deep research, incremental updates, and evolving focus profiles. Entry point for targeted topic intelligence.

Use when the user says 'scan', 'collect intel', 'run scan', or when invoked by cron. Orchestrates the full domain intelligence pipeline: collect from sources, filter duplicates, analyze insights, detect convergence signals, store results. Primary cron target.

2 hooks across 2 events