gate-info-web3

General Rules

⚠️ STOP — You MUST read and strictly follow the shared runtime rules before proceeding. Do NOT select or call any tool until all rules are read. These rules have the highest priority. → Read gate-runtime-rules.md → Also read info-news-runtime-rules.md for gate-info and gate-news shared rules (tool degradation, report standards, security, routing).

• Only call MCP tools explicitly listed in this skill. Tools not documented here must NOT be called, even if they exist in the MCP server.

CLI and playbook contract

1. Scope naming: Frame answers as Web3 / on-chain / protocol behavior. Do not narrow the narrative to “DeFi-only” unless the user’s question is strictly DeFi/TVL/yield. gate-info-defianalysis is a legacy alias for routing only — the canonical skill id is gate-info-web3.
2. Every CLI command MUST exist under gate-cli v0.5.2. Call only what appears in playbooks/gate-info-web3.yaml. When preflight route is CLI, do not invoke Gate MCP tools from this skill.
3. Always pass --format json on data-collection commands.
4. Separate evidence types in the report: label on-chain / CLI facts vs community or media interpretation (news, web search, UGC). Never present rumor as chain truth.
5. Required slots (address+chain, token+chain, platform, exchange+asset, symbol, entity_query) must be explicit; if ambiguous, ask — do not guess chain or contract.
6. Final user-facing reports may use the user’s locale; this SKILL.md stays English for discovery (see repository CLAUDE.md rule 1).

---

Step 0 — Preflight

Follow skills/_shared/preflight.md verbatim:

1. Run gate-cli preflight --format json; parse .route, .status, .user_message.
2. Branch: CLI → continue; MCP_FALLBACK → emit __FALLBACK__ and halt; BLOCK → echo user_message and halt.
3. When status == "ready_with_migration_warning", Step 3 appends one migrate hint at the end of the report.

Do NOT run any info / news data call until route == "CLI".

---

---

Step 1 — Intent routing

Map the user query to exactly one playbook id from playbooks/gate-info-web3.yaml.

Playbook id	When to pick	Required slots
address_tracking	Trace a wallet / “who is this address” / recent on-chain activity (behavior, not safety verdict).	address, chain
token_onchain	Token-level flow / holders / smart-money style signals for a known token + chain.	token, chain
entity_intel	Named desk / fund / entity (e.g. what a named desk is doing).	entity_query
protocol_platform	Protocol TVL, fees, volume, profile, history, or yield on a named platform (e.g. Uniswap, Aave).	platform
exchange_reserves	Exchange reserve / proof-of-reserves style questions for an asset.	exchange, asset
liquidation_heatmap	Perp / futures liquidation density by price (which price levels see liquidations).	symbol
stablecoin_bridge	Stablecoin market / peg / chain distribution, or bridge rankings — not a single-token DEX question.	—
token_onchain_social	User wants on-chain data plus news/sentiment/UGC in one pass.	token, chain

Slot extraction notes

• chain: normalize to ids the CLI accepts (eth, bsc, arbitrum, base, solana, ...). If missing, ask.
• platform: accept fuzzy names; if multiple matches are plausible, run gate-cli info platformmetrics search-platforms only with flags verified in gate-cli info platformmetrics search-platforms --help, pick one slug, then run get-defi-overview / get-platform-info / get-yield-pools as the playbook lists.
• token vs symbol: on-chain playbooks use token (+ chain); reserves / heatmap often use listed symbol (e.g. BTC, USDC).

Cross-skill routing at Step 1

User signal	Route to
Safety / honeypot / blacklist / compliance verdict (safety-first)	gate-info-risk (primary)
General research, fundamentals, macro, multi-coin compare without on-chain/protocol framing	gate-info-research (primary)
Pure “why did it move”, “community narrative”, no on-chain ask	gate-news-intel (primary)

---

Step 2 — Data collection

Execute the chosen playbook. Commands in the same parallel_group may run concurrently; wait for the group before the next. Only use commands defined in the playbook YAML (plus the optional search-platforms pre-check for fuzzy platform names when Step 1 says so).

Execution substitutes (v0.5.2)

• info +address-tracker / info +token-onchain: not relied upon. Use the address_tracking and token_onchain command blocks in the YAML (see cli_future_shortcut there for aspirational one-liners).
• info onchain trace-fund-flow and info onchain get-entity-profile: not in the v0.5.2 inventory documented in docs/gate-cli-commands-summary.md. Use get-address-info + get-address-transactions for flows; use news feed web-search for entity-style questions until get-entity-profile exists.
• News aggregates (news +brief, news +community-scan): use search-news, get-social-sentiment, search-ugc as in token_onchain_social.

Failure policy

• Required command fails → abort, surface CLI error, no fabricated section.
• Optional command fails → mark the matching subsection as no data in Step 3.
• News/web search may be partial or opinion-heavy — tag those lines as community / media, not as on-chain facts.

---

Step 3 — Synthesis

Produce exactly six sections, in this order (section titles in the final user report may be localized; structure is fixed):

Report template

## Web3 intel — {subject}

### 1. Subject and executive takeaways
- Objects (address / token / protocol / exchange / entity / theme)
- 1–3 neutral bullets; separate **on-chain verifiable** vs **needs confirmation**

### 2. On-chain activity (core)
- Recent interaction types, large transfers, label-based identity hints — **only** from on-chain / `info onchain` commands

### 3. Flows, positioning, smart money, entity read
- Flow and holder / positioning structure (`get-address-transactions`, `get-token-onchain`)
- Smart-money or whale-style signals (from token-onchain data)
- Entity questions: if `web-search` is used, mark as **second-hand** narrative

### 4. Protocol or platform metrics (if relevant)
- Scale (TVL, volume, fees), direction of change, pools or assets that matter (`platformmetrics` commands)

### 5. Community or news context (if relevant)
- Only when the playbook includes news / web-search: short factual + tone summary; **keep separate** from sections 2–3

### 6. Risks and watchlist
- Data gaps, follow-up on-chain signals, **non-advice** watch items

> This report is based on public data and tool output; it is not investment advice.

Migrate hint (only if Step 0 had ready_with_migration_warning)

Append verbatim the migration blockquote line from skills/_shared/report-style.md (the line starting with ⚙️ and gate-cli migrate --dry-run), same as other primary skills.

---

Cross-skill routing + Safety rules

Routing

Follow skills/_shared/routing.md. Web3-specific reminders:

Follow-up intent	Target
“Is it safe / honeypot / sanctioned?”	gate-info-risk (primary)
“Full research report / worth buying?” (no on-chain focus)	gate-info-research (primary)
“Why pump/dump / Twitter drama only”	gate-news-intel (primary)

Safety rules

Follow skills/_shared/report-style.md. Web3-specific:

1. No buy/sell recommendations; no price targets.
2. Do not treat empty on-chain data as proof of safety — that belongs in risk skill.
3. When web-search or UGC is used, prefix interpretive lines with media / community so they are not confused with chain state.

---

References

File	Load when...
skills/gate-info-web3/references/scenarios.md	Choosing among address / token / protocol / reserves / heatmap playbooks.
skills/gate-info-web3/references/cli-reference.md	Flag tables and verified platformmetrics / onchain / news flags.
skills/gate-info-web3/references/troubleshooting.md	Shortcut gaps, search-platforms discovery, optional get-transaction for a pasted tx hash.
skills/_shared/preflight.md	Step 0.
skills/_shared/routing.md	Cross-skill matrix.
skills/_shared/report-style.md	Shared report rules.