Build Cloudflare Worker orchestration layer for ERC20 + multi-agent ecosystem

[Copilot]

This PR adds a production-oriented Cloudflare orchestration layer that unifies token operations from Cf- with Theodore’s multi-agent endpoints behind a single authenticated Worker API. It also aligns infra/config/docs so the Worker, on-chain interactions, and CI deployment flow are wired end-to-end.

• Worker orchestration API (src/worker.js)

  • Added authenticated, CORS-enabled routes with uniform response envelope { success, data, timestamp }.
  • Implemented token routes:
    • POST /token/deploy (JSON-RPC deploy via eth_sendRawTransaction / eth_sendTransaction)
    • GET /token/status (name, symbol, totalSupply via eth_call)
    • POST /token/burn (burnTokens(amount) transaction flow)
  • Implemented agent routes:
    • POST /agent/finance
    • POST /agent/jobs
    • POST /agent/toolchain
    • GET /agent/status
  • Added orchestration triggers:
    • POST /cron (manual trigger)
    • scheduled handler (hourly pipeline execution and KV state updates)

• Agent modules (independently importable)

  • src/agents/finance.js: token analysis, market snapshot placeholder, KV snapshot store/read (1h TTL).
  • src/agents/jobs.js: job search placeholder, KV result store, D1 application read/write.
  • src/agents/toolchain.js: repo analysis suggestions, audit persistence/retrieval.
  • src/agents/tokenRpc.js: shared token ABI + JSON-RPC helpers used by Worker and finance agent.

• Cloudflare + data-plane configuration

  • Added wrangler.toml:
    • name = "creatingfire-orchestrator"
    • main = "src/worker.js"
    • KV bindings: AGENT_STATE, TOKEN_CACHE
    • D1 binding: THEODORE_DB
    • cron trigger: 0 * * * *
    • vars: NETWORK=mainnet
  • Added D1 schema at src/db/schema.sql:
    • job_applications
    • token_events
    • agent_runs
    • system_config

• Hardhat/runtime config updates

  • Expanded hardhat.config.js with dotenv, networks (mainnet, sepolia, polygon, localhost), gas reporter toggle, and etherscan key config.
  • Added per-network RPC env support (MAINNET_RPC_URL, SEPOLIA_RPC_URL, POLYGON_RPC_URL) with RPC_URL fallback.
  • Added worker scripts in package.json:
    • worker:dev
    • worker:deploy

• CI/CD + documentation

  • Added .github/workflows/deploy.yml for main:
    • test (npm test)
    • lint (npm run lint)
    • deploy-worker (wrangler deploy) gated on both checks and using Cloudflare secrets.
  • Replaced README with full architecture/operations doc:
    • system architecture
    • setup/deploy flow
    • Worker endpoint reference
    • environment variable matrix
    • CI/CD overview + repo links

Example route structure now exposed by the orchestrator:

export default {
  async fetch(request, env) { /* auth + route dispatch */ },
  async scheduled(_event, env, ctx) {
    ctx.waitUntil(runScheduledAgents(env, "scheduled"));
  },
};

[creatingfire2026]

Pilot security gate — do not merge or deploy this branch yet.

Required before reconsideration:

1. Remove automatic production deployment on every push to main. Use a staging environment first, GitHub Environment protection, and an explicit manual approval for production.
2. Change the default network from mainnet to a test network or local simulation. No token deployment or burn transaction may target mainnet during the pilot.
3. Do not expose a generic transaction relay. /token/deploy and /token/burn currently accept caller-supplied signed transactions and broadcast them without proving that the transaction matches the requested operation, expected chain, approved contract, sender, value, or calldata. Decode and validate every transaction, bind it to an allowlisted chain/contract/function, and add replay/idempotency protection.
4. Replace wildcard CORS with an explicit origin allowlist. Add rate limiting, request-size limits, structured input validation, audit identifiers, and secure error handling.
5. Separate read-only agent routes from privileged write/transaction routes. Use scoped credentials rather than one bearer secret for all capabilities.
6. Add tests for authentication failures, malformed input, RPC failures, transaction validation, replay attempts, wrong-chain requests, wrong-contract requests, and unauthorized state access.
7. Replace placeholder Cloudflare resource IDs and prove a complete staging deployment before production configuration is introduced.
8. Pin third-party GitHub Actions to immutable commit SHAs and add dependency/security scanning.

The architecture may continue as an isolated staging prototype, but production deployment and financial transaction execution remain outside the authorized pilot scope.