Agent Registry — Every Inc
Agent Registry
| Agent |
Directory |
Role |
Authority Scope |
Applicable Gates |
Human Owner |
| Editorial Quality |
editorial-quality/ |
Reviews articles against three rigor tests + AI tells detection |
Tier 1 (analysis), Tier 2 (detection), Tier 3 (publication) |
article-publication |
Kate Lee + Katie Parrott |
| Compound Engineering |
compound-engineering/ |
Executes Plan→Work→Review→Compound cycle |
Tier 1 (code gen/review), Tier 2 (auto-fix/compound), Tier 3 (merge) |
code-merge |
Product GM (Kieran/Naveen/Yash/Danny) |
| Consulting PM (Claudie) |
consulting-pm/ |
Manages consulting engagement logistics, status, deliverables |
Tier 1 (internal), Tier 2 (status/prep), Tier 3 (deliverables) |
consulting-deliverable |
Natalia Quintero |
| Content Distribution |
content-distribution/ |
Generates social posts from published articles |
Tier 1 (draft gen), Tier 2 (queue), Tier 3 (posting) |
social-media-publication |
Anthony Scarpulla |
| Product GM |
product-gm/ |
Full-stack specification authority: PRDs, compound engineering cycle, merge decisions |
Tier 1 (code gen/review/triage), Tier 2 (auto-fix/compound), Tier 3 (merge) |
code-merge |
Product GM (Kieran/Naveen/Yash/Danny) |
Directory Structure Per Agent
Each agent directory contains:
agents/{role-slug}/
system-prompt.md — Core: identity, context, boundaries, capabilities, collaboration, prompt, reporting
tool-permissions.md — Action/Permission/Condition table
self-review-checklist.md — Gate-derived checklist with checkbox items
Authority Tier Summary
| Tier |
Editorial Quality |
Compound Engineering |
Consulting PM (Claudie) |
Content Distribution |
Product GM |
| Tier 1 (Autonomous) |
Structural analysis, text scanning |
Code gen, test execution, 14-agent review |
Internal tracking, research, milestone monitoring |
Article extraction, voice profiling, draft generation |
PRD/plan writing, code gen, review pipeline, bug triage, research |
| Tier 2 (Autonomous + Notify) |
AI tells detection, rigor test evaluation |
Bug auto-fix, CLAUDE.md updates, knowledge base entries |
Status report generation, onboarding workflows, training prep |
Draft queuing for Anthony’s review |
P1 auto-fix, compound artifacts, knowledge base updates, design requests |
| Tier 3 (Human-in-Loop) |
Publication recommendation (Kate, 48h) |
Code merge (GM, 24h) |
Deliverable delivery (Natalia, 24h), client comms |
Social posting (Author + Anthony, 12h) |
Code merge (GM, 24h), cross-product data, shared infra |
| Tier 4 (Never) |
Editorial direction |
Architecture decisions, other GM’s config |
Financial commitments, cross-engagement data |
Social strategy |
Architecture decisions, pricing, other GM’s config |
Hard Boundary Coverage
| Boundary |
Editorial Quality |
Compound Engineering |
Consulting PM |
Content Distribution |
Product GM |
| 1. Never publish without review |
Primary |
– |
– |
Primary |
Relevant |
| 2. Never send external comms |
– |
– |
Primary |
Primary |
Relevant |
| 3. Never make financial commitments |
– |
– |
Primary |
– |
Relevant |
| 4. Never share client data cross-engagement |
– |
– |
Primary |
– |
– |
| 5. Never merge without review gate |
– |
Primary |
– |
– |
Primary |
| 6. Never change another GM’s config |
– |
Primary |
– |
– |
Primary |
| 7. Never exceed product PII scope |
– |
Primary |
– |
– |
Primary |
| 8. Never make ungrounded claims |
Primary |
– |
Primary |
Primary |
Relevant |
| 9. Never bypass quality gates |
Primary |
Primary |
Primary |
Primary |
Primary |
Future Agent Candidates
Based on the coordination audit’s encoding candidates:
| Candidate |
Domain |
Est. Hours Saved/Month |
Coordination Audit Reference |
| Design Request Intake Agent |
Design team |
~32 |
Quick Win #1: formalize design intake in Linear |
| Editorial Pipeline Agent |
Editorial |
~17 |
Quick Win #3: automated Kanban nudges |
| Cross-GM Knowledge Feed Agent |
Engineering |
~12 |
Quick Win #3: shared learnings from compound loops |
| Podcast Logistics Agent |
Podcast |
~12 |
Encoding candidate #5 |
Table of contents