A Pharos Skill is a structured folder that AI agents (Claude Code, Cursor, opencode, Windsurf, etc.) read at runtime. Each skill ships with a SKILL.md capability index, exact cast/forge command references, and battle-tested Solidity contracts — so your agent knows exactly what to run without guessing.

Skills

pharos-contract-verify   ──────────────────────────────── no deps
pharos-deploy-kit        ──→ pharos-contract-verify
pharos-token-factory     ──→ pharos-deploy-kit + pharos-contract-verify
pharos-safe-multisig     ──────────────────────────────── no deps
pharos-agent-wallet      ──────────────────────────────── no deps
pharos-x402-payments     ──→ pharos-agent-wallet

| Skill | What it does | |-------|-------------| | pharos-contract-verify | Verify any deployed Solidity contract on Pharos via Blockscout. Handles constructor args, compiler flags, and indexer retry. | | pharos-deploy-kit | Deploy via forge script or deterministic CREATE2/CREATE3 for same-address-across-networks. Dry-run gate before any broadcast. | | pharos-token-factory | Deploy ERC20/ERC721/ERC1155 tokens using audited OpenZeppelin v5 contracts. Mint, burn, pause, transfer ownership — with guardrails against dangerous combos. | | pharos-safe-multisig | Deploy and manage Gnosis Safe multisigs. Treasury management, owner changes, contract ownership transfer to Safe. | | pharos-agent-wallet | Autonomous wallet safety layer. Balance preflight, gas estimation, nonce management, per-session spend caps, and recipient allowlists. | | pharos-x402-payments | Full x402 HTTP micropayment stack. Monetize Express endpoints as a server and pay autonomously as a client — with spend caps and idempotency built in. |

Installation

npx pharos-skills add <skill>

The CLI prompts you to pick your editor(s) and install scope. Use arrow keys to move, space to select/deselect, and enter to confirm.

# Install a single skill (auto-resolves dependencies)
npx pharos-skills add pharos-token-factory

# Install all 6 skills at once
npx pharos-skills add-all

# Non-interactive / CI
npx pharos-skills add-all --yes

# Install into a specific directory
npx pharos-skills add pharos-token-factory --dir ./my-project

# Browse available skills
npx pharos-skills list

# Inspect a skill and its dependency tree
npx pharos-skills info pharos-x402-payments

What gets installed

<target>/
├── SKILL.md                        ← merged Capability Index for all installed skills
└── .pharos/
    ├── shared/
    │   ├── assets/
    │   │   ├── networks.json       ← Pharos testnet + mainnet RPC, chain IDs, explorers
    │   │   ├── tokens.json         ← canonical token registry
    │   │   └── canonical-contracts.json
    │   └── references/
    │       └── _guardrails.md      ← mandatory pre-check protocol for all write ops
    └── skills/
        └── <skill-name>/
            ├── SKILL.md
            ├── skill.json
            ├── references/        ← exact command specs per operation
            └── assets/            ← contracts, scripts, templates

Networks

| Network | Chain ID | RPC | Explorer | |---------|----------|-----|----------| | Atlantic Testnet (default) | 688689 | https://atlantic.dplabs-internal.com | atlantic.pharosscan.xyz | | Pacific Mainnet | 1672 | https://rpc.pharos.xyz | pharosscan.xyz |

[!IMPORTANT] Every skill defaults to Atlantic Testnet. Any mainnet write operation requires an explicit "This is a MAINNET transaction with real value" confirmation — this cannot be skipped.

Guardrails

Every skill enforces five non-negotiable safeguards:

1. Foundry pre-check — which cast && which forge before any command; prints install hint if missing.
2. Explicit private key — always --private-key $PRIVATE_KEY. Foundry does not auto-read env. Never hardcode, log, or echo keys.
3. 4-step Write Pre-Check — derive sender → confirm chain ID → verify balance covers value + gas → simulate before send. Cannot be skipped.
4. Testnet default + mainnet gate — all skills default to testnet; mainnet requires explicit confirmation.
5. Explorer link on every tx — <explorer>/tx/<hash> and <explorer>/address/<addr> always output after broadcast.

How an Agent Uses These Skills

1. Agent reads SKILL.md → scans the Capability Index for matching user intent
2. Opens the linked reference file (e.g., pharos-token-factory/references/erc20.md#deploy)
3. Follows the Agent Guidelines checklist at the bottom of each section
4. Runs Write Operation Pre-Checks from _guardrails.md
5. Executes exact cast/forge commands from the template
6. Outputs explorer links and records the result in a local ledger

Development

Prerequisites: Foundry · Node.js ≥ 20

# Install dependencies
npm install

# Build the CLI
npm run build

# Run CLI tests (22 tests, vitest 4)
npm test

# Compile Solidity contracts (solc 0.8.35 + OpenZeppelin v5.6.1)
cd contracts && forge build

# Run Solidity tests (15 tests)
cd contracts && forge test

# End-to-end install test
npx pharos-skills add pharos-token-factory --dir ./demo --yes

Contracts

All contracts are in contracts/src/, compiled with solc 0.8.35, via-ir, and 200 optimizer runs.

| Contract | Description | |----------|-------------| | StandardERC20.sol | ERC20 + Burnable + Pausable + Capped + Ownable | | StandardERC721.sol | ERC721 + URIStorage + Pausable + Ownable | | StandardERC1155.sol | ERC1155 + Pausable + Supply + Ownable | | SpendLimitGuard.sol | Gnosis Safe Guard — per-period cumulative spend cap |

[!NOTE] Contracts are designed as reference implementations. Audit before using in production.

Resources

• Pharos Network docs
• Foundry Book
• OpenZeppelin Contracts v5
• x402 Protocol
• Gnosis Safe docs