account-intel

Entity-centric footprint & reputation intelligence. Point the skill at one specific person, brand, company, or handle and it gathers paid signal across X/Twitter, YouTube, web news/reputation, web citations, and holdings from a user-supplied associated EVM wallet — over two payment protocols (x402 via Circle Gateway and MPP on Tempo) — which the agent fuses into a single cross-platform footprint brief: who this account is, how present they are per platform, and how credible the web makes them look.

When To Use

Use when the value is profiling a single entity — the question is who is this account, how present and credible are they across platforms. Examples: "profile @<handle>", "cross-platform footprint of <brand>", "reputation brief on <person>", "where does <handle> show up and is it legit".

This is distinct from social-intel, which does topic/keyword listening — sentiment and chatter on a subject across a crowd. If the user asks "what are people saying about <topic>" or "social sentiment on <subject>", use social-intel. If they name one account/brand/person and want that entity's footprint, use account-intel.

Rails

This skill declares two payment protocols and currently observes two routed modes:

• x402 via Circle Gateway / routed-x402 — the two SELAT-native X/Twitter reads, the Alchemy tokens-by-wallet read, and Exa web search.
• MPP on Tempo / routed-mpp — Scrape Creators YouTube search and Brave news search.

The manifest's catalog classification is mixed. Do not infer direct versus routed settlement from a provider hostname: selat skill verify --live-probe is the source of truth at execution time. All six steps probed as routed on 2026-10-04 (Exa now resolves routed-x402, not MPP); older reliability snapshots recorded some x402 steps as direct.

Workflow

1. Install: selat skill install account-intel
2. Collect all three inputs. handle, name, and address have no defaults — confirm the handle belongs to the named entity and that the user explicitly associates the wallet with it.
3. Verify all six live challenges for free: selat skill verify ~/.config/selat/skills/account-intel --handle <handle> --name "<entity>" --address <0x..> --live-probe
4. Show the user all six live prices and the expected total; get explicit approval and arm a cumulative session budget (selat budget start --amount <approved-cap>).
5. Run end-to-end, then disarm the budget (selat budget stop): selat skill run account-intel --handle <handle> --name "<entity>" --address <0x..>
6. The CLI compiles each step into a selat-pay call and prints each result.

Execution is a six-step pipeline. selat skill run always executes every manifest step and continues after an individual failure. It cannot stop early or omit Alchemy, so --address is required and must be a non-zero wallet the user explicitly associates with the entity. Supporting a truly optional on-chain step requires a separate skill or runner-level step selection.

The six manifest steps, in execution order (cheapest first; live quotes 2026-10-04):

1. X/Twitter profile — SELAT-native GET /twitter/user/info (x402 via Circle Gateway, ~$0.001) for follower counts, bio, verification.
2. X/Twitter recent tweets — SELAT-native GET /twitter/user/last_tweets (x402 via Circle Gateway, ~$0.001); read cadence + engagement, surface the breakout post.
3. On-chain wallet holdings — Alchemy POST /data/v1/assets/tokens/by-address (x402 via Circle Gateway, ~$0.001) across Ethereum, Base, Polygon (matic-mainnet), Arbitrum, and Optimism. Only meaningful when the user supplies a non-zero EVM wallet or treasury address that the user explicitly associates with the entity. Do not infer ownership from an untrusted search result.
4. Web citations — Exa POST /search (x402 via Circle Gateway, ~$0.007). Grounds the entity in indexed web sources for corroboration.
5. YouTube presence — Scrape Creators GET /v1/youtube/search (MPP on Tempo, ~$0.021) for whether/where the entity shows up on YouTube.
6. Web reputation / news — Brave POST /brave/news-search (MPP on Tempo, ~$0.03675); recent press, controversy, sentiment signal.

Then synthesize: a per-platform presence map (X, YouTube, web, on-chain), a credibility/reputation read, the entity's strongest channel, and where the web context confirms or contradicts the self-presentation — with source URLs.

Inputs And Outputs

Output: per-step JSON (X profile + recent tweets with engagement, YouTube search hits, Brave news results, Exa web results with text snippets + URLs, and Alchemy wallet holdings) that the agent fuses into a single entity footprint & reputation brief.

Gotchas

• Entity-centric, not topic-centric. This skill profiles one account/brand. For topic/keyword sentiment listening across a crowd, use social-intel.
• Router dependency is live, not hostname-derived. All six steps probed as routed on 2026-10-04. Require a reachable SELAT_ROUTER_URL for the current workflow and re-run verification if a mode changes.
• The social rails. The YouTube step pays Scrape Creators (mpp.orthogonal.com) via MPP on Tempo through the SELAT Router; the X/Twitter steps pay SELAT-native (catalog.selat.ai) through the SELAT Router.
• The on-chain step is operationally mandatory. selat skill run always executes Alchemy, so require a user-supplied non-zero address before quoting or paying. Never infer wallet ownership or substitute the zero address: the zero address can contain unsolicited balances and produce a large, misleading result.
• GET params in the query, POST params in body. SELAT-native and Scrape Creators are GET (?userName=/?query=). Brave, Exa, and Alchemy are POST; Alchemy requires an addresses array of { address, networks } objects; Polygon's Portfolio API network id is matic-mainnet (not polygon-mainnet).
• No silent defaults. Every param is required with no default, so a run never pays for a placeholder entity. Do not fill a missing handle or wallet with a well-known account.
• maxAmount is a per-step guardrail, not a cumulative run cap or price. Each step's cap is ~1.4–1.5x its live quote; caps sum to $0.0945. The manifest's top-level $0.05 is only a fallback for a step without its own cap — it is not a full-run cap, and every current step has an override. The cumulative limit comes from the armed session budget. Live quotes on 2026-10-04 (manifest order) were $0.001, $0.001, $0.001, $0.007, $0.021, and $0.03675, for an expected total of $0.06775.
• Paid failure can still cost money. The runner continues after a failed step. Check payment history before reporting; never auto-retry a charged failure without a fresh quote and approval.
• The live 402 is the source of truth. If a step stops serving a challenge, selat skill verify --live-probe flags it; the skill must be fixed (the step removed from the manifest) before it is run again.

Validation

--chain base in the probe commands below is only the flag selat-pay requires today — a probe reads a free, chain-independent quote and never settles. A real paid run resolves the settlement chain from your funded Circle Gateway balance, not the manifest.

• Static: selat skill validate ./skills/account-intel
• Live gate (free): selat skill verify ./skills/account-intel --handle <handle> --name "<entity>" --address <user-supplied-0x..> --live-probe
• Paid confirm (settles real 200s): only after an approved cumulative session budget, add --pay to that command and reconfirm the current total.
• Single-step probe (no pay). The example pairs Vitalik Buterin's handle with his publicly self-identified wallet (vitalik.eth); never substitute a token contract or an unrelated address as an entity's wallet: selat-pay GET "https://catalog.selat.ai/twitter/user/info?userName=VitalikButerin" --chain base --max-amount 0.0015 --probe-only --live-probe selat-pay POST "https://x402.alchemy.com/data/v1/assets/tokens/by-address" --body '{"addresses":[{"address":"0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045","networks":["eth-mainnet","base-mainnet","matic-mainnet","arb-mainnet","opt-mainnet"]}],"withMetadata":true,"withPrices":true,"includeNativeTokens":true,"includeErc20Tokens":true}' --chain base --max-amount 0.0015 --probe-only --live-probe

References

• manifest.json — the machine-readable payment recipe this skill runs.
• references/endpoints.md — the catalogue endpoints, rails, and live prices.
• references/agent-skill-authoring-sop.md — authoring standard.
• selat-pay — https://github.com/SELAT-AI/selat-pay

Third-party API names are trademarks of their respective owners; this skill only routes approved payments to their x402 and MPP endpoints.