story: Agent-experience pass — first-mile wallet page, x402 polish, troubleshooting examples, discoverability research

[GigaHierz]

Part of #2266.

User story

As a Web2 developer arriving for agent payments, I want to succeed on testnet without leaving the docs; as an AI agent, I want the troubleshooting facts the SDK docs leave out.

Context & evidence

Verified at bdf40b37:

• build-on-celo/build-with-ai/overview.mdx and x402.mdx start from AGENT_PRIVATE_KEY in an env var with no page answering "how does my agent get a wallet and test funds".
• x402.mdx uses facilitator without a plain-language definition, shows maxValue: "1000000" without saying what that is in dollars, and has no explicit success checkpoint. (The earlier "needs a thirdweb client ID first" blocker is already fixed — Celo's own facilitator is the featured path.)
• Troubleshooting cases that SDK docs cover in one paragraph — paying gas in USDC, adapter vs token address, 6 vs 18 decimals — are the examples worth keeping per the code-example policy in AGENTS.md (task: Add AGENTS.md — repo guidance for structure, headings, writing style, and code-example policy #2251).
• Agent-discoverability research exists and is to be folded in here once shared (OPEN: attach it to this issue).

Scope

1. One "Give your agent a wallet and test funds" page: key generation → faucet → first transfer; linked before every example that assumes AGENT_PRIVATE_KEY.
2. x402 page: define facilitator in one plain sentence at first use; inline-comment base units (maxValue: "1000000" = $1 for a 6-decimal token); an explicit "you just charged an agent $0.01" success checkpoint.
3. Troubleshooting examples across the agent pages: gas in USDC per SDK, adapter vs token address, decimals.
4. Fold in the discoverability research: actionable items done here or split into follow-ups.
5. Jargon pass on the agent journey: link ERC-20 / EVM on first use.

Acceptance criteria

• First-mile page live and linked from build-with-ai/overview and x402
• x402 changes as listed
• Research attached; each actionable item either done here or linked as a follow-up issue
• Every example run once; command and output in the PR

Non-goals

Self Agent ID (#2262); the AI-resources page (#2261); MiniPay (#2264).

Depends on / blocks

Depends on #2259 for final paths — content can be drafted earlier.

Measured at: bdf40b37