X/Twitter Research & Trust Network Mapping

Download

⬇ Download x-research.skill

Install with OpenClaw:

openclaw skill install x-research.skill

Or drop the file into your ~/.openclaw/skills/ directory.

What This Skill Does

This skill gives your OpenClaw agent the ability to research people and topics on X/Twitter by searching real tweets, not asking an LLM to guess. It covers:

• Account research — what does @someone actually say about a topic? Find real tweets with URLs.
• Trust network mapping — trace who recommends what to whom. Follow reply chains to discover accounts and products you'd never find through search.
• Method selection — bird CLI vs Playwright vs Grok API, with honest benchmarks on speed, accuracy, and failure modes.
• Cross-referencing — combine X findings with Reddit community consensus for stronger signal.
• Product discovery — from trusted endorsements, not ads. Verify stores, check prices, trace the recommendation chain.

When Your Agent Uses It

The skill triggers when you ask your agent to:

• Research what specific X accounts think about something
• Find product recommendations from people you trust
• Map who endorses whom in a community
• Cross-reference claims across multiple accounts
• Verify whether someone actually said something on X

The Key Insight: Reply Chains > Direct Search

The most valuable technique isn't searching what someone says — it's searching what others recommend to them.

When a trusted account asks "looking for a good shampoo," the replies are a curated recommendation list from their community. That's higher-signal than any search engine result because:

• The recommenders self-selected (they follow the account, share values)
• The recommendations are specific (real products, real experiences)
• You can trace the recommender's credibility (check their profile too)

# The magic query: what do people recommend TO an account?
bird search "to:TrustedAccount recommend OR try OR use"

# Then trace each recommender
bird about RecommenderAccount
bird search "from:RecommenderAccount topic"

Method Comparison

The Grok API warning is critical. In testing, Grok confidently invented example tweets, fabricated product endorsements, and got an account's stance on DHT blockers exactly backwards. It has a training cutoff and no real-time X access. Never use it to determine what someone actually said on X.

What's in the Package

x-research/
├── SKILL.md                              # Core instructions, method selection, search patterns
└── references/
    └── trust-network-example.md          # Full worked example with real data

Requirements

• OpenClaw — any recent version
• bird CLI — npm install -g bird — X/Twitter API access via ct0 cookie auth
• Playwright (optional) — pip install playwright && playwright install chromium — for profile scraping and store verification
• xAI API key (optional) — for Grok queries on non-X topics only

Background

This skill was built from a real research session: finding the best shampoo by mapping the trust network of eight health influencer accounts on X. The process discovered FountainHead Hair — a brand that never would have appeared in a Google search — by tracing reply chains from @Tallowtwins' community.

Along the way, we tested every available method side-by-side: bird CLI, manual Playwright scrolling, X Search, browser Grok, and the Grok API. The results were stark — bird CLI outperformed everything else by a wide margin, and Grok API actively produced harmful misinformation.

The trust network methodology generalizes beyond product recommendations. It works for finding developers, evaluating projects, sourcing information, or mapping any community where who-endorses-whom matters more than what-ranks-highest.

Freedom Tech Perspective

Traditional product research routes you through affiliate-driven review sites, SEO-optimized listicles, and paid placements. The trust network approach inverts this — you start from people you actually trust and follow the signal.

• No algorithmic curation — you choose the seed accounts, not a recommendation engine
• Verifiable chains — every claim links to a real tweet URL you can check
• Anti-hallucination by design — bird CLI returns real data; the skill explicitly warns against Grok's fabrication tendency
• Runs locally — bird CLI and Playwright run on your machine. The xAI API is the only external call, and it's optional
• Web of Trust model — the same concept behind PGP key signing and Nostr verification, applied to product recommendations

This is how research worked before search engines: you asked people you trusted, who pointed you to people they trusted. The tools just make it faster.