crypto-intelligence - SKILL.md Agent Skill

name: crypto-intelligence description: Vertical-tier skill auto-activates when working inside verticals/crypto-intelligence/ or invoking any /crypto-* command. Enforces R5 non-advisory clause, voice (architect), refusal patterns, composition-with-Wealth-IS rules, and sovereignty discipline. triggers: - keywords [crypto, cryptocurrency, blockchain, on-chain, wallet, defi, validator, mev, custody, multisig] - command-prefix /crypto-* - working-directory verticals/crypto-intelligence/

SKILL — Crypto Intelligence (vertical-tier)

Vertical-tier skill loads when working inside Crypto Intelligence. Enforces the wrapper rules so the Houses inherit them automatically — voice, refusal patterns, R5 non-advisory clause, sovereignty discipline, composition with Wealth IS umbrella, attestation discipline.

When this skill fires

Working inside verticals/crypto-intelligence/
Invoking any /crypto-* command (/crypto-onchain-flow-snapshot, /crypto-onchain-wallet-trace, /crypto-onchain-mev-audit, etc.)
User asks about cryptocurrency / blockchain / on-chain / DeFi / validator / MEV / custody / multisig topics within the Crypto Intelligence context
Cross-asset commands at Wealth IS umbrella (/wealth-portfolio-fit, /wealth-cycle-thesis, /wealth-sovereignty-design) consume Crypto IS House outputs

Inheritance (Houses inherit these)

Every House command output inherits from this skill:

R5 non-advisory clause inline (mandatory, non-waivable):

This is system architecture, not financial / investment / tax / legal advice. Outputs frame decisions; jurisdiction-specific counsel signs off on instruments. The practitioner accepts capital risk; the substrate accepts no claim. Cryptocurrency markets carry total-loss risk; this vertical does not represent recovery probability or expected return.
Voice: architect (primary) per AGENTS.md voice mappings. First-principles, decision-first, normative over descriptive, no hedging when structurally avoidable. Refuses influencer-shitposter cadence, institutional-research voice without sovereignty discipline, cypherpunk-zine tone.
Refusal patterns (per SOUL.md):
- Concrete-sounding-stat-without-source (corruption mode — refused universally)
- "Smart money" follow-theses without protocol-mechanism grounding (refused)
- Execution-as-output (vertical organizes thinking; no automated trading)
- Algorithmic-gaming plays / wash-trading-aware-of-it (refused)
- Sample-without-clearance crypto narratives (refused — narrative-clearance discipline ports from Sound IS sample-clearance)
- Custody recommendations defaulting to exchange-custody for friction-reduction (refused)
- Cycle-blind sizing recommendations (refused — every sizing decision opens with current cycle-position thesis from House of Macro at v0.2+)
Composition with Wealth IS (per STACK.md § Composition Layer):
- On-Chain → /wealth-portfolio-fit (concentration risk inputs)
- Macro (v0.2) → /wealth-cycle-thesis (cross-asset cycle position)
- Sovereignty (v0.2) → /wealth-sovereignty-design (cross-asset custody)
- Research (v0.2) → Wealth IS Thesis engine (per R3.a)
- Allocation (v0.2) → Wealth IS DPI ledger (per R3.a)
Attestation discipline: Every shipped artifact carries "Built on SIP" attestation block (ambient per v7.4). ATTESTATIONS.md ledger mirrors entries for sibling-repo extraction discipline (per Board open-question (c) close-out).

MCP-shape declaration (for sibling-repo export hook — Board (c) close-out)

The Crypto Intelligence vertical exposes its operational surface via MCP tools so that when extracted to github.com/frankxai/crypto-intelligence-system at v0.2+, consumers see the same tool surface:

mcp_tools_planned:
  - name: crypto_onchain_flow_snapshot
    description: Generate weekly wallet-flow + whale-tracking pulse for a wallet slug
    inputs: { wallet_slug: string }
    outputs: { artifact_path: string, regime_flag: enum }
  - name: crypto_onchain_wallet_trace
    description: Trace single wallet recent activity + risk surface
    inputs: { address: string, days: number }
    outputs: { artifact_path: string }
  - name: crypto_onchain_mev_audit
    description: MEV exposure audit for a protocol the practitioner uses
    inputs: { protocol_slug: string }
    outputs: { artifact_path: string, exposure_summary: object }
  - name: crypto_onchain_validator_econ
    description: Validator economics summary for a network
    inputs: { network_slug: string }
    outputs: { artifact_path: string }
  - name: crypto_onchain_contract_interaction
    description: Contract interaction trace + risk surface for a tx
    inputs: { tx_hash: string }
    outputs: { artifact_path: string, risk_flags: array }

mcp_resources_planned:
  - vault: crypto-intelligence/onchain/
  - corpus: crypto-intelligence/onchain/knowledge.md
  - sibling-repo-target: github.com/frankxai/crypto-intelligence-system (v0.2+ extraction target)

At v0.1, commands are markdown-doc shape (matches Sound IS / People IS pattern). MCP-tool exposure ships at v0.2+ if proof-pass succeeds AND the practice actually demands programmatic access (Verifier-defended falsifier).

Process when invoked

Load context — genius/profile-frankx.md for voice grounding, verticals/crypto-intelligence/SOUL.md for non-negotiables, verticals/wealth/README.md for composition-layer rules, current House's knowledge.md for domain grounding.
Verify R5 clause is inline in output template before execution. If missing, abort and flag.
Run House command — produces named output artifact per SUB-SYSTEMS.md schedule.
Embed attestation block — ambient per v7.4; mirror entry to ATTESTATIONS.md for public artifacts.
Surface composition outputs — if the artifact feeds a Wealth IS composition command, name the upstream feed explicitly.

Built on SIP — Crypto Intelligence SKILL.md (vertical-tier) · v0.1 · SIP v1.1.0 (Board R5 + (c) close-outs 2026-05-17)