arena-research-skill（竞技场研究技能）

Arena Research

General-purpose agentic research over Are.na. Decompose any research question into targeted searches, explore channels, follow the connection graph to discover adjacent ideas, identify key curators, deep-dive linked content, and synthesize into a sourced briefing.

For API details (endpoints, auth, response format): read references/arena-api.md.

When to Use Arena vs Other Sources

• Are.na — curated references, visual research, design thinking, cultural analysis,

reading lists, how people organize ideas. High signal, human-curated, long-tail.

• X/Twitter (x-research) — real-time reactions, dev discourse, product launches,

breaking news, expert hot takes. Fast-moving, high volume.

• Web search — factual answers, documentation, current events, specific URLs.

Are.na is a library, not a firehose. The value isn't just what people saved — it's *how they organized it* and *what else they connected it to*.

CLI Tool

All commands run from this skill directory:

cd ~/clawd/skills/arena-research
source ~/.config/env/global.env

Search

bun run arena-search.ts search "<query>" [options]

Options:

• --type Channel|Block|Text|Image|Link|User|Group — filter result type (default: all)
• --sort score|created|updated|connections|random — sort order (default: score)
• --scope all|my|following — search scope (default: all, my/following require auth)
• --per N — results per page (default: 24, max: 100)
• --page N — page number
• --quick — quick mode: 10 results, 1hr cache, channels only, sort by connections
• --save — save to ~/clawd/drafts/arena-research-{slug}-{date}.md
• --json — raw JSON output
• --markdown — formatted markdown output

Examples:

bun run arena-search.ts search "tools for thought" --type Channel --sort connections
bun run arena-search.ts search "brutalist web design" --quick
bun run arena-search.ts search "cybernetics" --type Link --per 50
bun run arena-search.ts search "spatial computing" --scope my

Channel

bun run arena-search.ts channel <slug-or-id> [options]

Options:

• --sort position|created|updated — content sort (default: position)
• --type Text|Image|Link|Attachment|Embed|Channel — filter content type
• --per N / --page N — pagination
• --connections — show channels that share blocks with this one (graph traversal)
• --save / --json / --markdown

Examples:

bun run arena-search.ts channel arena-influences
bun run arena-search.ts channel arena-influences --type Link --per 50
bun run arena-search.ts channel arena-influences --connections

Block

bun run arena-search.ts block <id> [options]

Options:

• --connections — show which channels this block appears in (graph traversal)
• --json

Examples:

bun run arena-search.ts block 3235876
bun run arena-search.ts block 3235876 --connections

User

bun run arena-search.ts user <slug-or-id> [options]

Options:

• --per N / --page N
• --json

Me

bun run arena-search.ts me

Shows authenticated user's profile and channels. Requires ARENA_ACCESS_TOKEN.

Cache

bun run arena-search.ts cache clear

Research Loop (Agentic)

When doing deep research (not just a quick search), follow this loop:

1. Decompose the Question into Search Strategies

Turn the research question into 3-5 search queries approaching the topic from different angles.

Think about the topic at multiple levels of abstraction:

• Direct terms: The obvious keywords (spatial computing, brutalist web design)
• Adjacent concepts: Related fields (haptic interfaces, concrete architecture)
• Practitioner language: How Are.na users actually name things (tools for thought,

digital gardens, vernacular web)

• Broader category: The umbrella topic (interaction design, web aesthetics)
• Specific references: Known works, people, or projects in the space

Start with channels sorted by connections to find the most-connected collections first:

bun run arena-search.ts search "tools for thought" --type Channel --sort connections --per 20

2. Search and Assess Channels

Run each query. After each search, assess:

• Which channels have real depth? Look at item counts in results. Channels with

2-3 items are stubs. Channels with 50+ items are serious collections.

• Who owns them? If the same user or group appears across multiple relevant channels,

they're a key curator worth exploring directly.

• Are there channels I should explore in full? Pick the top 3-5 by content count

and relevance.

Also search for blocks directly to find specific linked content:

bun run arena-search.ts search "cybernetics" --type Link --sort connections --per 20

3. Explore Top Channels

For each promising channel, fetch its contents:

bun run arena-search.ts channel tools-for-thought --per 50

To see only external links (the most valuable for research deep-dives):

bun run arena-search.ts channel tools-for-thought --type Link --per 50

Look for:

• Link blocks with source URLs — these are external references worth deep-diving
• Text blocks with content — these are notes, annotations, and original writing
• Image blocks — note titles and descriptions, deprioritize for text-based research
• Nested channels (type: Channel) — sub-collections worth exploring

4. Follow the Connection Graph

This is what makes Are.na research uniquely powerful. Two strategies:

4a. Block connections — "Where else does this content live?"

For high-signal blocks (interesting links, well-described text), check which other channels they appear in:

bun run arena-search.ts block 3235876 --connections

A block in 28 channels means 28 different people thought it was worth saving. The channels it appears in show you how different people contextualize the same idea.

4b. Channel connections — "What's adjacent to this collection?"

For a channel you've identified as high-quality, find other channels that share content:

bun run arena-search.ts channel arena-influences --connections

This surfaces *conceptual neighbors* — channels that share blocks with the one you're exploring. A channel about "tools for thought" might connect to "cybernetics," "memory palaces," and "personal knowledge management."

Graph depth: One hop is the sweet spot. Block→connections→those channels is enough. Two hops gets exponential fast. Stop at one hop unless you have a specific reason to go deeper.

5. Identify Key Curators

As you explore, track which users appear repeatedly:

• Same user owns multiple relevant channels
• Same user connected blocks to several channels you've explored
• Users with high follower counts in the space

Fetch their profile and channels:

bun run arena-search.ts user charles-broskoski

6. Deep-Dive Linked Content

When blocks are Link-type and have source URLs, use web_fetch to read the actual content. Prioritize links that:

• Appear in many channels (high connection count from step 4a)
• Come from channels with many items or followers
• Point to essays, blog posts, GitHub repos, research papers, or documentation
• Are directly relevant to the research question

Skip deep-diving:

• Social media posts (use x-research)
• Image galleries or portfolios
• Paywalled content
• Dead links (check the state field — available means it's live)

7. Search Your Own Collections

If looking for something you've personally saved:

bun run arena-search.ts search "query" --scope my

Or content from people you follow:

bun run arena-search.ts search "query" --scope following

8. Synthesize

Group findings by theme, not by search query. Each theme should capture a conceptual cluster of channels, blocks, and curators.

### [Theme/Finding Title]

[1-2 sentence summary of what curators are collecting and how they frame it]

**Key channels:**
- [Channel Title](https://www.are.na/owner/slug) by @username — N items
  [Brief description of scope and quality]
- [Channel Title](https://www.are.na/owner/slug) by @username — N items
  [Brief description]

**Notable content:**
- [Block title](source_url) — found in N channels
  [Why it's significant]

**Key curators:**
- [@username](https://www.are.na/slug) — N channels, N followers
  [What they focus on]

**Connected territory:**
- [Adjacent channel](https://www.are.na/owner/slug) — overlaps via shared blocks
  [What this connection reveals about the topic]

9. Save

Use --save flag or save manually to ~/clawd/drafts/arena-research-{topic-slug}-{YYYY-MM-DD}.md.

Include a metadata footer:

---
## Research Metadata
- **Query**: [original question]
- **Date**: YYYY-MM-DD
- **Source**: Are.na v3 API
- **API calls**: N search queries + N channel fetches + N block connection lookups + N deep-dives
- **Channels explored**: N
- **Blocks scanned**: ~N
- **Search terms used**: [list the actual search strings]
- **Limitations**: [any gaps]

Refinement Heuristics

• Too many shallow channels? Filter for items > 10 in results
• Too few results? Broaden keywords, try synonyms, drop --type filter
• Want the most-saved content? Search blocks with --sort connections
• Looking for specific link types? Use --type Link for external references, --type Text for notes
• Channel too large to browse? Use --type and --sort on the channel command
• Want to spider outward? Pick 3 blocks with most connections, fetch their connections, see what emerges
• Stuck? Search for a known reference or thinker, find their channels, explore from there
• Visual research? Use --type Image — Are.na is heavily used for mood boards and art direction

Content Types Quick Reference

File Structure

skills/arena-research/
├── SKILL.md              (this file)
├── arena-search.ts       (CLI entry point)
├── lib/
│   ├── api.ts            (Are.na v3 API wrapper: search, channels, blocks, users)
│   ├── cache.ts          (file-based cache, 15min TTL)
│   └── format.ts         (terminal + markdown formatters)
├── data/
│   └── cache/            (auto-managed)
└── references/
    └── arena-api.md      (Are.na v3 API endpoint reference)