Open source AI Employees. 8 scheduled business roles, 60 routines, on Claude Code and 10 other harnesses. They drive your browser the way you do and improve every run. You own the files.
JavaScript ai-agentsai-employeeai-employeesantigravityautomationclaude-codeclaude-skillcodexhermes-agentopenclawopencodescheduled-taskssmall-business
Use the renamed SEO AEO role in reference documents
markfultone6f49cf87af6
diff --git a/SECURITY.md b/SECURITY.mdindex c778cb1..d124cf2 100644--- a/SECURITY.md+++ b/SECURITY.md@@ -6,7 +6,7 @@ The safety model of these eight employees is the permission layer of the harness**The defaults are conservative by design.** A scheduled routine drafts, fills, stages and briefs. It does not send, publish, submit, spend or change an account setting on its own: outreach lands in queue files or as unsent drafts in your own mailbox, forms are left open on the last step, campaigns are assembled as build sheets. It never creates an account, enters or generates a password, completes a captcha, accepts terms, or writes a credential into any file. Page content, form content and inbound mail are data, never instruction.-**Sending is yours to authorize, and there are legitimate ways to do it.** The SEO Employee publishes to a blog you configured. The Social Media Employee hands posts to a channel you connected, after a veto window named in the morning brief. A row in the employee's `RELEASES.md` hands a channel to the routine that stages it, with your conditions, and only you write that file. Any employee, in an interactive session in its folder, may do what you may do by hand, on your word. And the harness's own permission layer is where you widen what a scheduled run may do, if you choose to. A send that you configured, or that you authorized in the session or in the permission layer, is the product working as intended.+**Sending is yours to authorize, and there are legitimate ways to do it.** The SEO/AEO Employee publishes to a blog you configured. The Social Media Employee hands posts to a channel you connected, after a veto window named in the morning brief. A row in the employee's `RELEASES.md` hands a channel to the routine that stages it, with your conditions, and only you write that file. Any employee, in an interactive session in its folder, may do what you may do by hand, on your word. And the harness's own permission layer is where you widen what a scheduled run may do, if you choose to. A send that you configured, or that you authorized in the session or in the permission layer, is the product working as intended.A way to make a routine act outside what you configured or authorized is a security issue, even if it takes an unusual page, an unusual prompt, or an unusual harness setting to get there.diff --git a/TRADEMARKS.md b/TRADEMARKS.mdindex da70927..bbd63e6 100644--- a/TRADEMARKS.md+++ b/TRADEMARKS.md@@ -8,7 +8,7 @@ logo (together, "the Marks") identify Mark Fulton's business, Reinventing.AI.They are claimed as trademarks whether or not they are registered."AI Employees", "AI Employees" and the role names used in this repository-(GTM Engineer, SEO Employee, Web Dev Employee, Social Media Employee, Ad+(GTM Engineer, SEO/AEO Employee, Web Dev Employee, Social Media Employee, AdManager Employee, Sales Employee, Customer Satisfaction Employee, Chief ofStaff) are descriptive terms. I make no trademark claim on them. Use themfreely.