Driver
The marketing slide "Treat Claude Code Like A Team" presents 7 personas as equal team members. Framework reality:
- 2 have proper names: code-reviewer is "Rex", security-reviewer is "Shield"
- 5 roles have only titles: Product Manager, Tech Lead, Backend Engineer, Frontend Engineer, QA Engineer
- 3 other agents have NO names either: dependency-auditor, pr-manager, ticket-manager
The naming is inconsistent: not all agents have names (3 don't), and no roles have names. The marketing implies all 7 are equivalent named personas; the code says only 2 are.
This creates two problems:
- Adopter confusion — adopters reading the marketing slide expect to invoke "Pat the QA Engineer" or "Maya the PM" and find no such handle anywhere in the framework.
- Inconsistent agent UX — when running
/audit-deps, the dependency auditor's output is unsigned; when running /code-review, Rex's output is signed. Same shape of tool, different visibility.
Scope
Pick one approach and apply it consistently.
Option A — Give every role + agent a proper name
Add a persona_name field (or similar) to every role file in roles/{dept}/{role}.md and every agent file in .claude/agents/*.md. Examples:
| Role / Agent |
Proposed name |
| Product Manager |
Maya |
| Tech Lead |
Luca |
| Backend Engineer |
Alex |
| Frontend Engineer |
Jamie |
| QA Engineer |
Pat |
| Code Reviewer (Rex) |
unchanged |
| Security Reviewer (Shield) |
unchanged |
| Dependency Auditor |
??? (TBD) |
| PR Manager |
??? (TBD) |
| Ticket Manager |
??? (TBD) |
The marketing slide updates to show all names. When a role activates, the main agent introduces itself with the persona name.
Option B — Drop names from Rex and Shield, use titles uniformly
Rename code-reviewer agent's introduction from "I'm Rex" to "I'm the Code Reviewer". Same for Shield → "Security Reviewer". Marketing slide updates to remove the "(REX)" / "(SHIELD)" labels.
Recommendation
Option A. Names create memorable identity, ease team-feel framing the marketing already invested in, and adopters can refer to them as "Pat said the AC isn't met yet" — natural team-language. Drop-the-names is technically simpler but loses the brand equity already built around Rex/Shield.
Acceptance Criteria
Risks / Dependencies
- Naming bikeshed — picking specific names is opinion-territory. Recommendation: file one AgDR with the names + reasoning, ship it, move on. Adopters can override per-fork via
roles/*/ edits if they want their own names.
- Marketing slide drift — the slide is a separate artifact; when this lands, slide 5 needs an update PR. Track as a linked task in this ticket.
Refs
- Surfaced during marketing-vs-reality audit (mid-session in atlas-apex's adoption test).
- Related: role-activation visibility ticket (sibling — being filed alongside this).
Driver
The marketing slide "Treat Claude Code Like A Team" presents 7 personas as equal team members. Framework reality:
The naming is inconsistent: not all agents have names (3 don't), and no roles have names. The marketing implies all 7 are equivalent named personas; the code says only 2 are.
This creates two problems:
/audit-deps, the dependency auditor's output is unsigned; when running/code-review, Rex's output is signed. Same shape of tool, different visibility.Scope
Pick one approach and apply it consistently.
Option A — Give every role + agent a proper name
Add a
persona_namefield (or similar) to every role file inroles/{dept}/{role}.mdand every agent file in.claude/agents/*.md. Examples:The marketing slide updates to show all names. When a role activates, the main agent introduces itself with the persona name.
Option B — Drop names from Rex and Shield, use titles uniformly
Rename
code-revieweragent's introduction from "I'm Rex" to "I'm the Code Reviewer". Same for Shield → "Security Reviewer". Marketing slide updates to remove the "(REX)" / "(SHIELD)" labels.Recommendation
Option A. Names create memorable identity, ease team-feel framing the marketing already invested in, and adopters can refer to them as "Pat said the AC isn't met yet" — natural team-language. Drop-the-names is technically simpler but loses the brand equity already built around Rex/Shield.
Acceptance Criteria
Risks / Dependencies
roles/*/edits if they want their own names.Refs