1inch (Script-First, Install-and-Use)

This skill is designed for your architecture: do not rely on platform-injected tools.

• ✅ Uses only skill-local scripts under skills/1inch/scripts/
• ✅ Agent executes scripts via bash (python3 ...)
• ✅ No oneinch_* tool calls required

Why this version fixes “tool not found”

Old design depended on runtime tool registration (oneinch_quote, oneinch_swap, ...). If tool injection fails, agent is blocked.

New design uses deterministic local scripts:

1. call 1inch HTTP API directly
2. call wallet service directly (OIDC via /.fly/api)
3. print JSON result

So after install, the agent always has a runnable path.

Files

• scripts/_oneinch_lib.py — shared client + wallet/OIDC helpers
• scripts/tokens.py — token search/list
• scripts/quote.py — quote only
• scripts/check_allowance.py — allowance check
• scripts/approve.py — approve tx broadcast
• scripts/swap.py — swap execution (optional auto-approve)
• scripts/run_swap_flow.py — one-command flow (quote + optional approve + swap + post-trade balance verification)

Environment

Required:

• ONEINCH_API_KEY
• WALLET_SERVICE_URL (default fallback exists)
• sc-proxy connectivity for 1inch API (mandatory)

Assumptions:

• Running on Fly Machine with /.fly/api unix socket for OIDC token minting.

sc-proxy requirement (critical)

This skill enforces 1inch calls through sc-proxy.

• It reads proxy from HTTP_PROXY/HTTPS_PROXY first, else falls back to PROXY_HOST + PROXY_PORT.
• If no proxy env is found, scripts fail fast with a clear error instead of silently direct-connecting.

Supported chains

ethereum, arbitrum, base, optimism, polygon, bsc, avalanche, gnosis

Agent execution rules (IMPORTANT)

When user asks for 1inch actions, follow this exact pattern:

1. Never call oneinch_* tools
2. Run local scripts with bash + python3 skills/1inch/scripts/<script>.py ...
3. Parse JSON output and respond with concise summary
4. For write actions, always do post-check using wallet balance tools (if available) or rerun script-based checks

Canonical mapping

• “查 token 地址 / 列 token” → tokens.py
• “先报价 / 看能换多少” → quote.py
• “检查授权” → check_allowance.py
• “授权” → approve.py
• “执行兑换” → swap.py

Command examples

1) Search token

python3 skills/1inch/scripts/tokens.py --chain polygon --search POL --limit 10

2) Quote: 1 USDC -> POL

python3 skills/1inch/scripts/quote.py --chain polygon --from USDC --to POL --amount 1

3) Check allowance (USDC)

python3 skills/1inch/scripts/check_allowance.py --chain polygon --token USDC

4) Approve USDC (unlimited)

python3 skills/1inch/scripts/approve.py --chain polygon --token USDC

5) Swap: 1 USDC -> POL (auto-approve if needed)

python3 skills/1inch/scripts/swap.py --chain polygon --from USDC --to POL --amount 1 --slippage 1.0 --auto-approve

6) One-command full flow (recommended)

python3 skills/1inch/scripts/run_swap_flow.py --chain polygon --from USDC --to POL --amount 1 --slippage 1.0 --auto-approve

Tune retries when needed:

python3 skills/1inch/scripts/run_swap_flow.py --chain polygon --from USDC --to POL --amount 1 --slippage 1.0 --auto-approve --swap-retries 3 --swap-retry-backoff 2

Default workflow for “买 X USDC 的 POL”

Preferred deterministic sequence:

1. run_swap_flow.py --auto-approve
2. Read JSON result and report quote, tx submission result, and verification deltas

Fallback manual sequence (if user asks step-by-step):

1. quote.py (confirm expected output)
2. swap.py --auto-approve
3. Verify with fresh balances (before/after)

If swap returns wallet policy rejection:

• load wallet-policy skill
• propose wildcard baseline (DENY exportPrivateKey, ALLOW *)
• after user confirms, rerun swap command

Error handling

• Unknown chain → ask user to choose supported chain
• 1inch API 4xx/5xx → show raw error; run_swap_flow.py automatically retries transient /swap 5xx (default: 2 retries with exponential backoff)
• Not enough <token> balance from 1inch → likely symbol mapped to a different token variant (e.g. USDC.e vs USDC); rerun with explicit token address from tokens.py
• Wallet API 4xx/5xx → show raw error, do not fabricate tx hash
• insufficient allowance (without auto-approve) → rerun with --auto-approve or run approve.py
• policy rejection → propose policy update then retry

Notes

• Amount input is human units (e.g. --amount 1 = 1 USDC), script handles decimal conversion.
• Token symbol resolution comes from 1inch /tokens on the selected chain.
• If a symbol has multiple variants on a chain (e.g., USDC / USDC.e), prefer passing token contract address explicitly to avoid ambiguity.
• For native token input, use native / ETH.

Quick smoke test

python3 skills/1inch/scripts/quote.py --chain polygon --from USDC --to POL --amount 1

If this works, the skill is operational in script mode.

Script Reference (CLI args)

This skill exposes scripts (not Python functions). Run them via bash + python3 skills/1inch/scripts/<name>.py <args>. All scripts emit JSON on stdout — parse with json.loads(...) or just print the output.

Common arguments

Scripts

Recommended flow

For ANY user "swap X for Y" request, default to run_swap_flow.py --auto-approve. It handles the full chain end-to-end and verifies via wallet balance read after broadcast.

If user asks "just quote, don't execute" → quote.py only.

Output schema (run_swap_flow.py)

JSON with these top-level keys:

• quote — {fromToken, toToken, fromAmount, toAmount, estimatedGas, ...}
• approve — {required, txHash, success} (only if --auto-approve)
• swap — {txHash, success, error?}
• verification — {from_balance_before, from_balance_after, to_balance_before, to_balance_after, delta_in, delta_out}

If wallet returns a policy rejection, swap.success = false and swap.error will mention policy. Load the wallet-policy skill, propose a wildcard baseline, then rerun.