AGENTS.md — SLV Repository (OSS)

⚠️ THIS IS AN OPEN-SOURCE REPOSITORY

Every file, commit, PR, and comment is publicly visible. Treat every character you write as if it will be read by the entire world — because it will be.

Security Rules (MANDATORY — NO EXCEPTIONS)

🚫 NEVER include in any file, commit message, PR, or comment:

• API keys, tokens, secrets, passwords, or credentials
• Internal or private API endpoints
• Internal hostnames, IPs of private infrastructure
• Authentication headers (e.g., Authorization: Bearer ...)
• Private environment variable values
• Internal database URLs or connection strings
• References to internal tools, dashboards, or monitoring systems
• Customer data, user IDs, or personally identifiable information

✅ Safe to include:

• Publicly documented API endpoints
• Public documentation URLs (erpc.global, docs, GitHub)
• Generic placeholder values (e.g., YOUR_TOKEN_HERE, <server-ip>)
• Ansible playbook usage with generic examples
• Architecture descriptions without internal implementation details

External Communication

• HTTP requests in code: Any fetch(), curl, or HTTP call MUST use only public endpoints
• Tokens in code: Use placeholder constants (e.g., EPRC_ZERO_BLOCK), never real credentials
• Environment variables: Reference by name only ($SOLANA_RPC_URL), never embed actual values
• Webhook URLs, callback URLs: Never hardcode internal URLs

Before Every Commit

1. grep -r 'Bearer ' --include='*.ts' --include='*.yml' — check for leaked tokens
2. grep -r 'erpc.global' --include='*.ts' --include='*.yml' — verify only publicly documented endpoints are referenced
3. Review any new HTTP calls — ensure they target only public endpoints
4. If any check fails → DO NOT COMMIT. Remove the reference first.

Repository Overview

SLV is an open-source toolkit for Solana validator and RPC node deployment and management.

• CLI: cli/ — Deno-based CLI tool (slv command)
• Ansible: template/{version}/ansible/ — Deployment playbooks
• Jinja: template/{version}/jinja/ — Configuration templates
• Skills: dist/oss-skills/ — Self-contained AI agent skill packages

Development Flow

1. Branch from main → implement → PR
2. All PRs require review before merge
3. Automated tests must pass
4. Security scan on every PR (no secrets, no internal refs)

Skill Packages (dist/oss-skills/)

Each skill is a self-contained package with:

• SKILL.md — AI-readable playbook knowledge + interactive deployment flow
• AGENT.md — Agent persona and behavior rules
• ansible/ — Playbooks with resolved dependencies
• jinja/ — Configuration templates
• examples/ — Sample inventory files

These are designed to be installed by AI coding agents (OpenClaw, Codex, Claude Code, etc.) to enable automated Solana infrastructure management.

Bot Templates (slv bot init)

Available templates in ValidatorsDAO/solana-stream → temp-release/:

Trade App (trade-app)

PumpSwap (Pump.fun AMM) new pool detection and auto-trading bot.

Lifecycle: Pool detected → Buy → TX confirm → Sell monitor → Sell → Drain → ATA close → Profit notification

Required env vars (.env):

• GRPC_ENDPOINT — Geyser gRPC endpoint
• SOLANA_RPC_ENDPOINT — Solana RPC (reads)
• SOLANA_SEND_RPC_ENDPOINT — Solana RPC (TX send, separate recommended)

Optional:

• WEBHOOK_URL — Discord notifications
• REDIS_URL — Trade history persistence
• API_TOKEN — API authentication

Trade config (PUT /api/config):

Setup:

slv bot init          # → select "trade-app"
cd ~/slv/<app-name>
cp .env.sample .env   # → configure env vars
cargo build --release
./target/release/trade-app
# Fund wallet with at least 0.013 SOL
curl -X POST http://localhost:3000/api/trade/start

API: OpenAPI docs at http://localhost:3000/docs

Discord notifications: ✅ Buy Confirmed / 🟢🔴 Trade Complete / ⚠️ Retreat Burn

Warning: wallet.json contains your private key. Never commit it.

See temp-release/trade-app/README.md for full details.