Research Intelligence Processor

Process a link through the full intelligence pipeline. Read config.json at the project root for business context.

Input

A URL to research: /research <URL>

Before Starting

1. Read config.json to get the business name, categories, and voice settings
2. Read data/research/intelligence-brief.md to understand existing signals and context
3. Read data/portfolio/projects.md to know what projects are active

Checkpointing

This skill uses the intake-log status as a checkpoint. Update the status after each major step so progress is visible and recoverable:

On failure: The intake-log status shows exactly where it stopped. To resume, check the status and pick up from that step. Do NOT re-run completed steps.

Pipeline

Step 1: Log

Add to the TOP of data/research/intake-log.md:

`[YYYY-MM-DD HH:MM]` | pending | Title (@author if known) | URL

Step 2: Fetch & Understand

For tweets/X posts:

• Use WebFetch on the URL
• If the tweet references an article, repo, or video — fetch THAT too
• If the tweet has a thread, try to get the full thread

For articles:

• Use WebFetch to read the full article
• Extract: thesis, key claims, methodology, data points, tools mentioned

For videos (YouTube, X video, etc.):

• Get title, description, duration from the page
• Mark as pending-transcript — the USER generates transcripts
• Create placeholder in data/knowledge/transcripts/ with metadata

For GitHub repos/tools:

• Fetch README, check stars, recent activity
• Create evaluation in data/knowledge/tools/
• Do NOT install tools — add to user's task list

For strategies/guides:

• Read and extract the full methodology
• Create a file in data/knowledge/strategies/

Checkpoint: Update intake-log status to step-2/9 before fetching. After successful fetch, update to step-3/9.

Step 3: Analyze for Business Applicability

Read the business categories from config.json and evaluate the content through EACH lens. For every category, answer: "What does this mean for the business? What should they DO?"

If a category doesn't apply to this content, say so briefly and move on.

Checkpoint: After analysis, update to step-4/9.

Step 4: Update Knowledge Base

• If this is a deep methodology or strategy: create a file in data/knowledge/strategies/
• If this is a tool evaluation: create a file in data/knowledge/tools/
• If it's a transcript: create placeholder in data/knowledge/transcripts/

Checkpoint: After knowledge base updates, update to step-5/9.

Step 5: Update Docs

• Add to the appropriate topic section in data/research/bookmarks.md
• If this reveals a new signal or changes an existing one, update data/research/intelligence-brief.md
• If the author produces consistently valuable content, add to data/research/people-to-watch.md

Checkpoint: After docs updated, update to step-6/9.

Step 6: Report

Show the user a structured summary:

## Research: "Title"

**Source:** domain.com | Date
**Relevance:** HIGH/MEDIUM/LOW — one sentence why

### Analysis

**[Category 1]:** What this means + what to do
**[Category 2]:** What this means + what to do
...

### Actions Added
- [ ] Specific action item (Project: which project)

### Updated Files
- file1.md (what changed)

Step 7: Mark Processed

Change the intake-log entry from pending to processed.

Step 8: Cross-Project Check

Read data/portfolio/projects.md. Does this research affect any active project? If yes, add a recommendation to the Recommendations table.

Step 9: Log Ideas

If the research suggests something worth building, adopting, or offering — add it to data/portfolio/implementation-backlog.md with status idea.

Checkpoint: After backlog updated, change intake-log status to actioned.

Content Access Workarounds

When encountering inaccessible content, use these before marking as user task:

When you can't access something: Log it to data/research/intelligence-brief.md with what the content is, why it matters, what was attempted, and what user action is needed.

Division of Labor

Claude does: Fetch, read, analyze, document, maintain all research files, create action items. User does: Generate transcripts, install/test tools, engage on social, create published content.

Principles

• Depth over speed — understand the content fully before analyzing
• Business lens — every insight must connect to an action
• Honest assessment — don't hype things that aren't ready
• Connect dots — link to existing signals and knowledge
• Follow the thread — if they reference something, fetch it
• Track people, not just content — add high-value producers to people-to-watch.md

ARGUMENTS: url