Add autonomous agent manager, agent API, business generation helpers, and CSV hardening

[bry92]

Motivation

• Introduce an autonomous business agent to run periodic automation steps for businesses and expose control endpoints.
• Factor out LLM-driven business generators from the business router to improve reuse and testability.
• Harden CSV export to properly quote cells and neutralize spreadsheet formula injection payloads.

Description

• Add AIBusinessAgentManager (server/_core/aiBusinessAgent.ts) implementing start/stop/runNow, state tracking, logs, and a singleton aiBusinessAgentManager for runtime agent orchestration.
• Add tests for the agent manager (server/_core/aiBusinessAgent.test.ts) that exercise start, runNow, and stop flows.
• Add an agent router (server/routers/agent.ts) with start, stop, status, and runNow endpoints that use executeAutonomousStep to generate or optimize business assets via existing database helpers and the LLM.
• Extract LLM-driven generation logic into server/routers/businessGenerators.ts providing generateBrandingForBusiness, generateWebsiteForBusiness, generatePricingForBusiness, and generateLeadsForBusiness, and update businessRouter to call these helpers instead of embedding LLM prompts.
• Register agentRouter in server/routers.ts.
• Improve CSV export (server/export.ts) by introducing escapeCsvCell and makeCsvRow to always quote cells and prefix suspicious values that start with =,+,-,@,\t,\r with a single quote to neutralize formula injection.
• Update CSV tests in server/export.test.ts to expect quoted headers/cells and to check that formula injection payloads are neutralized.

Testing

• Ran unit tests with vitest including server/_core/aiBusinessAgent.test.ts and the updated server/export.test.ts.
• The new agent unit test executed start, manual runNow, and stop flows and passed.
• The updated export tests validating quoting, escaping, and formula-neutralization ran and passed.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a03e477896

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Prevent overlapping cycles when restarting an active agent

Calling start on an already-running agent only clears the interval timer, but it does not cancel or gate an in-flight runCycle before immediately launching a new one. If a user triggers start twice in quick succession (or retries after a timeout), two executions for the same business can run concurrently, which can duplicate LLM calls and race the read-then-insert generation steps (get*Result + create*Result) into duplicate records.

Useful? React with 👍 / 👎.