my-project-public

repository

loading code, commits, and activity

repositories

loading repo index

#1	# Solana Clawd Agent Creation Guide
#2
#3	## Overview
#4
#5	Agents on Solana Clawd are specialized AI assistants designed to handle specific tasks across the Solana DeFi, trading, NFT, and SPL-token ecosystem. This guide shows you how to create, publish, and deploy agents into our hub at [/agents](https://www.x402.wtf/agents) so other builders can install and run them.
#6
#7	Every agent you publish becomes:
#8
#9	- 🪐 Discoverable via our JSON API and manifest
#10	- 🔌 Installable into any MCP client (Clawd Desktop, Cursor, ClawdOS, etc.)
#11	- 🎨 Mintable as an on-chain MPL Core asset on Solana (optional, via `/agents-mint`)
#12	- 🗣️ Addressable through our A2A (agent-to-agent) endpoint at `/api/agents/a2a`
#13	- 📈 Trackable in the hosted registry at `/agents-registry`
#14
#15	---
#16
#17	## Agent Anatomy
#18
#19	### Basic Structure
#20
#21	```json
#22	{
#23	"author": "your-github-or-solana-handle",
#24	"config": {
#25	"systemRole": "Your detailed system prompt here...",
#26	"openingMessage": "Optional welcome message",
#27	"openingQuestions": ["Suggested first question?"]
#28	},
#29	"identifier": "unique-agent-id",
#30	"meta": {
#31	"title": "Agent Display Name",
#32	"description": "Clear, concise description (max 200 characters)",
#33	"avatar": "🤖",
#34	"tags": ["solana", "defi", "trading"],
#35	"category": "defi"
#36	},
#37	"schemaVersion": 1,
#38	"createdAt": "2026-04-16"
#39	}
#40	```
#41
#42	### Key Components
#43
#44	identifier: URL-safe name (lowercase, hyphens, no spaces). Used as the JSON filename and in API routes.
#45
#46	- Good: `solana-jupiter-router`, `pump-fun-sniper`
#47	- Bad: `My Agent!`, `Pump.Fun Agent v2`
#48
#49	title: Display name shown in [/agents](https://www.x402.wtf/agents).
#50
#51	description: Short summary shown in the agent card (160–200 chars optimal).
#52
#53	avatar: Single emoji that represents the agent's function.
#54
#55	tags: 3–8 keywords. Common Solana tags: `solana`, `spl`, `jupiter`, `raydium`, `pumpfun`, `meteora`, `drift`, `marginfi`, `kamino`, `nft`, `mpl-core`, `metaplex`, `priority-fees`, `mev`, `validator`, `stake`.
#56
#57	category: One of `defi`, `trading`, `nft`, `analytics`, `security`, `dev-tools`, `education`, `governance`.
#58
#59	systemRole: The core prompt that defines agent behavior.
#60
#61	---
#62
#63	## Writing Effective System Prompts
#64
#65	### Structure Template
#66
#67	```text
#68	You are [ROLE/IDENTITY on Solana]
#69
#70	CAPABILITIES:
#71	- [What the agent can do]
#72	- [Key specialization — which Solana programs, DEXs, or SDKs]
#73	- [Tools/knowledge available]
#74
#75	GUIDELINES:
#76	- [How to approach tasks]
#77	- [Response format]
#78	- [Constraints/limitations — e.g. priority fee ceilings, slippage caps]
#79
#80	WORKFLOW:
#81	1. [Step-by-step process]
#82	2. [How to handle specific scenarios]
#83
#84	EXAMPLES:
#85	[Optional: Demonstrate desired behavior]
#86	```
#87
#88	### Solana DeFi Agent Example
#89
#90	```json
#91	{
#92	"systemRole": "You are a Solana Yield Optimization Specialist focused on the Kamino, MarginFi, Drift, and Meteora ecosystems.\n\nCAPABILITIES:\n- Analyze yield opportunities across lending, LP, and staking protocols on Solana\n- Calculate net APY factoring in borrow rates, LP fees, and reward emissions\n- Monitor priority fees and suggest optimal transaction timing via Helius/Triton RPC\n- Compare concentrated liquidity ranges on Meteora DLMM vs Orca Whirlpools\n\nGUIDELINES:\n- Always include risk assessments (high/medium/low) with reasoning\n- Factor priority fees (lamports/CU) and Jito tips into net return math\n- Flag programs with recent audits and known incidents\n- Never promise outcomes; this is analysis, not financial advice\n\nWORKFLOW:\n1. Ask for capital size, risk tolerance, and time horizon\n2. Check current market conditions (SOL price, funding rates, TVL moves)\n3. Compare 3–5 opportunities with pros/cons table\n4. Provide specific Jupiter routes or program instructions\n5. Include warnings on smart contract, oracle, and peg risks\n\nFORMAT:\nProtocol \| APY \| TVL \| Risk \| Priority Fee \| Recommendation"
#93	}
#94	```
#95
#96	---
#97
#98	## Best Practices
#99
#100	### Do's
#101
#102	✅ Be specific about which Solana programs/DEXs the agent understands
#103	✅ Use accurate Solana terminology (CU, priority fees, Jito tips, PDAs, CPIs)
#104	✅ Provide structured, scannable output (tables, bullets)
#105	✅ Mention audit status (OtterSec, Neodyme, Sec3, Halborn) when relevant
#106	✅ Include disclaimers about rug risks on pump.fun / new SPL tokens
#107	✅ Use clear step-by-step workflows referencing real tool calls
#108
#109	### Don'ts
#110
#111	❌ Make financial guarantees or price predictions
#112	❌ Conflate Solana with EVM concepts (there's no "gas" — use "priority fees" / "compute units")
#113	❌ Ignore program audit status or track record
#114	❌ Recommend interacting with unverified programs without warnings
#115	❌ Copy prompts from other agents verbatim
#116
#117	---
#118
#119	## Agent Categories on Solana Clawd
#120
#121	### 🪙 DeFi & Finance
#122
#123	Jupiter routing, Kamino vaults, MarginFi lending, Drift perps, Meteora DLMM, Orca Whirlpools, Sanctum LST, stablecoin peg monitoring.
#124
#125	### 📈 Trading & Analysis
#126
#127	Pump.fun sniping, Raydium LP analysis, Phoenix orderbook making, MEV/sandwich detection, whale tracking via Helius webhooks, on-chain flow via Bitquery/Flipside.
#128
#129	### 🎨 NFT & MPL Core
#130
#131	Tensor/Magic Eden floor analysis, MPL Core asset minting, compressed NFT (cNFT) workflows, royalty enforcement, collection bootstrapping.
#132
#133	### 🛠️ Dev & Security
#134
#135	Anchor program auditing, Solana Playground helpers, Squads multisig review, IDL explainers, priority-fee estimators, transaction simulation via LiteSVM/Mollusk/Surfpool.
#136
#137	### 📚 Research & Education
#138
#139	Solana fundamentals, SVM internals, validator economics, stake delegation, governance (Realms, SPL-Governance), tokenomics for SPL launches.
#140
#141	---
#142
#143	## Testing Your Agent
#144
#145	Before submitting to the [/agents](https://www.x402.wtf/agents) hub, test thoroughly:
#146
#147	1. Role adherence: Does it stay on Solana topics?
#148	2. Edge cases: How does it handle a non-Solana question?
#149	3. Error handling: Does it gracefully decline out-of-scope tasks?
#150	4. Response quality: Accurate compute unit estimates? Correct program IDs?
#151	5. Consistency: Same tone and format across conversations?
#152
#153	### Test Prompts
#154
#155	```text
#156	"What's the best SOL-staking yield right now, net of fees?"
#157	"Compare providing liquidity on Meteora DLMM vs Orca Whirlpool for SOL/USDC"
#158	"Calculate impermanent loss for a ±5% DLMM range on JLP/USDC"
#159	"Is program ID <PROGRAM_ID> safe to interact with?"
#160	"Explain how Jito tips affect transaction landing rates"
#161	```
#162
#163	---
#164
#165	## Localization
#166
#167	Agents are automatically translated into 18 languages when merged. Tips for translation-friendly prompts:
#168
#169	- Use clear, standard English
#170	- Avoid idioms and colloquialisms
#171	- Keep sentences concise
#172	- Use consistent terminology (don't alternate "SOL" and "Solana native token")
#173	- Define technical terms on first use
#174
#175	---
#176
#177	## Submission Checklist
#178
#179	- [ ] Unique, descriptive identifier
#180	- [ ] Clear title and description
#181	- [ ] Appropriate avatar emoji
#182	- [ ] 3–8 relevant Solana tags
#183	- [ ] Category assigned
#184	- [ ] Well-structured system prompt with real Solana references
#185	- [ ] Tested with at least 10 prompts
#186	- [ ] No copyrighted or proprietary content
#187	- [ ] JSON validates correctly (`bun run build`)
#188
#189	---
#190
#191	## Publishing to the Solana Clawd Hub
#192
#193	Three ways to publish:
#194
#195	### 1. PR into the repo (recommended)
#196
#197	```bash
#198	git clone https://github.com/clawdsolana/OpenClawd.git
#199	cd openclawd/defi-agents
#200	cp agent-template.json src/your-agent-name.json
#201	# edit your agent
#202	bun run format && bun run build
#203	# open a PR
#204	```
#205
#206	Once merged, your agent appears at:
#207
#208	- `https://www.x402.wtf/agents` (gallery)
#209	- `https://www.x402.wtf/{your-agent-name}.json` (API)
#210	- 18 localized variants auto-generated
#211
#212	### 2. Mint as on-chain MPL Core asset
#213
#214	Visit [/agents-mint](https://www.x402.wtf/agents-mint) to register your agent as a Metaplex Core asset. This gives it a transferable Solana identity, optional royalty rails, and an entry in the on-chain [/agents-registry](https://www.x402.wtf/agents-registry).
#215
#216	### 3. Host externally, register via A2A
#217
#218	Expose your agent over our A2A protocol at `POST /api/agents/a2a` (see [API.md](./API.md)) and submit the registration endpoint — your agent stays on your infra but is discoverable through the hub.
#219
#220	---
#221
#222	## Examples of Great Agents
#223
#224	### Simple Agent (Solana Sentiment)
#225
#226	```json
#227	{
#228	"identifier": "solana-sentiment-analyzer",
#229	"meta": {
#230	"title": "Solana Sentiment Analyzer",
#231	"description": "Analyzes Solana ecosystem sentiment from X, Discord, and on-chain flow",
#232	"avatar": "📊",
#233	"tags": ["solana", "sentiment", "analysis", "social"],
#234	"category": "analytics"
#235	},
#236	"config": {
#237	"systemRole": "You analyze Solana ecosystem sentiment by synthesizing X (Twitter) trends, Discord chatter, validator governance votes, and on-chain flow (netflows to CEX, LST mint/burn rates, new token creations on pump.fun). Provide sentiment scores (bullish/neutral/bearish) with evidence and confidence levels."
#238	}
#239	}
#240	```
#241
#242	### Complex Agent (Solana Portfolio Manager)
#243
#244	```json
#245	{
#246	"identifier": "solana-portfolio-manager",
#247	"meta": {
#248	"title": "Solana Portfolio Manager",
#249	"description": "Multi-protocol position tracking, rebalancing, and risk management across Solana DeFi",
#250	"avatar": "💼",
#251	"tags": ["solana", "defi", "portfolio", "risk", "kamino", "marginfi"],
#252	"category": "defi"
#253	},
#254	"config": {
#255	"systemRole": "You are an expert Solana portfolio manager specializing in multi-protocol position tracking across Kamino, MarginFi, Drift, Meteora, Orca, and Sanctum.\n\nCORE FUNCTIONS:\n1. Portfolio Analysis: Index positions via Helius DAS API, compute net worth in SOL and USD\n2. Risk Assessment: Liquidation thresholds on lending, IL exposure on LP positions, LST depeg risk\n3. Rebalancing: Suggest optimal allocation based on market regime and user goals\n4. Yield Tracking: Monitor APYs, reward claim timing, compound strategies, and Jito MEV rewards\n\nWORKFLOW:\n- Request Solana pubkey or manual position entry\n- Fetch positions via Helius / Birdeye\n- Surface risk exposures with specific liquidation prices\n- Provide actionable recommendations with Jupiter routes\n- Include priority-fee estimates and recommended Jito tips\n\nOUTPUT FORMAT:\nPortfolio Summary \| Risk Score \| Recommendations \| Action Items\n\nAlways disclaim: 'Not financial advice. DYOR.'"
#256	}
#257	}
#258	```
#259
#260	---
#261
#262	## Need Help?
#263
#264	- Open an issue in [clawdsolana/OpenClawd](https://github.com/clawdsolana/OpenClawd/issues) with the `agent-help` label
#265	- Review the [Prompt Engineering Guide](./PROMPTS.md)
#266	- Check existing agents in [`defi-agents/src/`](../src/) for inspiration
#267	- Read [DEPLOYMENT.md](./DEPLOYMENT.md) for hosting options
#268
#269	---
#270
#271	Ready to contribute? Open a PR, and your agent goes live on [x402.wtf/agents](https://www.x402.wtf/agents) within 24 hours.
#272

z6Mkq5mY3JWtxoxUobWcfNHm7AkRubgSWEZTkBVqZXJviFZ5/my-project-public