Autonomous cross-chain AI agent orchestration framework with tri-node architecture: Base (social/identity), Arbitrum Stylus (risk compute), Solana (execution). Exposes blockchain state tools to LLM for real-time on-chain awareness.
Three tools with complete input schemas and descriptions, but significant gaps in parameter constraints, output documentation, and error handling. Tool names follow verb_noun convention (read_*, run_*, push_*), which is good. However, descriptions lack actionable context for LLM selection, parameters are under-constrained (no enums, ranges, or format specs), and output schemas are not documented. Error handling returns JSON but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.
Send a notification or status update to a user via their Farcaster Frame on Base. Used for reporting trade execution results, risk rejections, or portfolio updates.
Fetch current liquidity and price data for a token pair from Jupiter Aggregator on Solana. Returns best routes, available liquidity, and estimated price impact.
Execute a risk simulation on the Arbitrum Stylus contract. Returns a risk score (1-100) based on current market volatility and the user's risk tolerance. Higher scores indicate riskier strategies.
No output schemas documented. LLMs cannot predict response structure or plan downstream calls. read_solana_liquidity returns JSON with 'success', 'data', 'error' fields, but this is inferred from code, not declared in tool definition.
Parameters lack constraints. 'volatilityIndex' and 'userTolerance' accept any number; no enum for tolerance (1|2|3), no range validation (0-100 for volatility). 'amount' unbounded. LLMs will pass invalid values.
No tool annotations. read_solana_liquidity and run_arbitrum_risk_sim are read-only; push_farcaster_notification is destructive (sends external message). Missing readOnlyHint/destructiveHint prevents agents from reasoning about side effects and retry safety.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | <=2025-11-25 | v2 |
Error responses lack recovery guidance. handleReadSolanaLiquidity returns {success: false, error: msg} but does not suggest next steps (e.g., 'Try with a different token pair' or 'Check Jupiter API status'). Agents cannot self-correct.
Descriptions lack context for LLM selection. 'Fetch current liquidity...' does not explain WHEN to call this vs. other price sources, or what prerequisites exist (e.g., 'Requires valid Solana token mints'). Descriptions average ~65 chars; baseline is 194 chars for A+ tools.