Price / promo drift — propagate every Mexico surface same turn
How an AI ops team runs this job end-to-end — agents do the grunt work, you approve anything public, paid, or irreversible.
The job: WC ⇄ dashboard ⇄ ManyChat sheet ⇄ guías ⇄ social plates. International stays independent. Show the checklist, not the dollars.
Open source · MIT
Agentic Company
Agents crawl, summarize, draft assets, and prepare publish/sync. You approve anything public, paid, or irreversible.
View on GitHub →
The wish (say this once)
Run this job on a cadence. Crawl or pull data, analyze, draft the artifacts, then stop for my OK before publish / deploy / send.
Phases
- Pull — APIs, crawls, dashboards (no invented numbers in public copy)
- Analyze — themes, priorities, effort tags
- Draft — assets, pages, captions, Open Tasks
- Gate — stop before live social, FTPS, paid ads, customer mail
- Ship — execute the approved action
Recipe refs: Job 3, Recipe #7.
Hard gates (always)
- Ask before any paid API spend (quote USD first).
- Local demo URL before production deploy.
- Approve & post for social; no customer emails in test probes.
- Public posts: no absolute company $, no customer PII — blur screenshots.
Install the same job
The open playbook lives in Agentic Company — recipes, integrations, and the autonomy model.
Closing
ANDAFACIL is a case-study pattern, not a data dump. Builders: start from the repo, not from someone else’s KPIs.
— Alex
Open source · MIT
Agentic Company
Agents crawl, summarize, draft assets, and prepare publish/sync. You approve anything public, paid, or irreversible.
View on GitHub →
