Remaining T3 user-facing branding

[gannonh]

Status

Approved

Rewritten 2026-08-25 from a flat punch list into this epic. Phase 1 (#109) shipped core Kata identity and deferred the leftovers below. Do not build this parent. Build approved children only.

Goal

Clear leftover T3 naming from surfaces a Kata Code user, operator, or contributor still reads, without renaming durable formats or violating FORK.md.

Context

Related: #109 (Phase 1 identity, closed), #119 (internal identifiers, wire formats, storage keys — not this epic), #124 (vendor-pull; do not stall it on this epic).

Audit 2026-08-25: docs/user/updating.md already uses npx @kata-sh/code-cli@latest (and versioned specs). Most other user-visible t3@ / npx t3 / t3 pair strings remain. The previous body mixed those with marketing, legal, telemetry, MCP wire names, and theme ids.

FORK.md still lists apps/marketing/** copy, t3.json, packaging/aur/**, and .repos/** as intentionally T3-shaped.

Constraints and non-goals

• Do not rename t3.json, Connect wire IDs, checkpoint refs, OAuth URNs, T3_* env vars, or t3code: client storage keys. That is #119.
• Do not rewrite apps/marketing/** copy, legal pages, store listing IDs, or tweets until Kata release URLs, store apps, and legal entity exist.
• Do not rename MCP server id t3-code, OTLP service name t3-server, theme ids t3-chat / t3-code, or EAS owner: pingdotgg in a child that is not explicitly about that rename.
• Do not invent winget / Homebrew / AUR package names. Native packages are not published for Kata yet.
• Historical path matchers may still recognize t3 / t3@ cache layouts so old installs keep being detected.

Acceptance criteria

• #127 (update copy, pairing labels, install/docs) is status:verified.
• Every remaining deferred area below is either a verified child or an explicit skip comment on this issue that records the product decision.
• No child of this epic takes work owned by #119.
• FORK.md leftovers (apps/marketing/**, AUR, t3.json) stay T3-shaped until a child is approved that overrides that policy.

Architecture

User-visible copy and docs first. Wire names, telemetry, themes, and marketing stay parked until a named decision. Later slices are independent user outcomes (each with its own blocker), not a backend-then-frontend waterfall.

#118 remaining user-facing branding (this epic)
  ├─ #127 update/reconnect copy, PTY error, pairing labels, install + operator docs
  ├─ later: GitHub feedback routing (templates, CONTRIBUTING)
  ├─ later: MCP / provider clientInfo t3-code (lockstep; may break existing MCP configs)
  ├─ later: OTLP / trace service names (dashboard break)
  ├─ later: marketing, legal, store listings (entity + listings)
  ├─ later: mobile EAS owner (Expo account)
  └─ later: theme ids (prefs migration vs display-label-only)
#119 internal identifiers (sibling, not a child)

Delivery slices

1. Update copy, pairing labels, and install docs — #127. Per-surface required names live on that child (versioned npm spec vs CLI verb vs npx hint).
2. Feedback routing. #146 keeps issue templates and CONTRIBUTING.md inside the Kata repository.
3. MCP / provider identity — lockstep t3-code to a Kata name across adapters + tests. Blocked until we decide whether existing user MCP configs must keep working under the old name.
4. Telemetry names — t3-server, t3-headless-relay-client, t3.client.surface. Blocked until dashboards can follow the rename.
5. Marketing / legal / stores — blocked on Kata release URLs, App Store / Play listings, and legal entity.
6. Mobile EAS owner — blocked on the Expo account that should own builds.
7. Theme ids — blocked on migrate-with-alias vs display-label-only.

Demonstration

Delegated to children. This epic is verified when every child is verified or explicitly skipped.

Verification

Per child. Program-level: each child's Verification grep and name table pass on the surfaces that child owns. Do not treat this parent as greppable scope.

Risks and mitigations

• Punch-list gravity. The old body invited one PR for everything. Children stay one concern; this parent stays draft until remaining slices are decided or skipped.
• Vendor-pull collision. #124 slice 2+ will re-import T3 copy. Remap to Kata names in those ports; do not wait for this entire epic.
• False internals. #119 owns durable keys and T3_* env vars. If a string is both user-visible and a storage key, stop and split rather than rename the key here.

Build handoff

• Approved scope: none on this parent. Build the first approved child only.
• Non-goals: marketing, legal, telemetry keys, MCP wire names, theme ids, EAS, Retain upstream-compatible internal T3 identifiers #119 internals, vendor-pull slices.
• Ordered slices: copy/docs first; everything else waits on a product call or a later child spec.
• Required verification commands: per child.
• Fixtures or credentials needed: none for slice 1.
• Blocking open questions: slices 3–7 as listed. Slice 1 (Kata names in update copy, pairing labels, and install docs #127) has none.

[gannonh]

Progress assessment (2026-08-25)

Verdict: not done. Phase 1 (#109) shipped core Kata identity on product surfaces and intentionally deferred most of this punch list. A repo audit today shows ~1/10 sections complete, with scattered partial fixes in a few files.

Done

• Docs — updating.md — all stale npx t3@… commands replaced with npx @kata-sh/code-cli@… (lines 35, 43, 50, 70).

Partial (file touched, checklist items still open)

• selfUpdate.ts — user-facing repair hint now says katacode service install, but version-skew strings at :80 and :167 still say t3 / t3@.
• cli/service.ts — CLI output says "Kata Code service" but status/install lines still print t3@<version>.

Still open (verified in tree)

Section	Status
Server update / version-skew copy	6/6 locations still t3@ or "t3 version"
Error copy (BunPtyAdapter)	still npx t3
Session labels (pair/connect/project)	still t3 pair, t3 connect cli, t3 project cli
Telemetry service names	still t3-server, t3-headless-relay-client, t3.client.surface, docs examples
Provider/MCP identity	still t3-code across all adapters + CodexDeveloperInstructions
Marketing site	still pingdotgg/t3code, T3 Code copy, T3 store IDs, legal pages
Docs — install.md, server-updates.md, observability.md	upstream release links, winget/brew/AUR package names, npx t3, $HOME/.t3 trace paths
Mobile EAS owner	still pingdotgg
Themes	still t3-chat / t3-code ids (needs migration vs display-label decision)
Feedback routing	issue templates + CONTRIBUTING still point at pingdotgg/t3code discussions

Likely confusion

This issue is easy to mix up with #109 (Phase 1 identity cut), which is largely complete and explicitly left marketing copy, theme ids, MCP wire names, and many user-visible strings for later sweeps like this one.

Suggested next slices (priority order)

1. Server update copy + BunPtyAdapter + session labels — small, user-visible, low risk.
2. Docs (install, server-updates, observability) — broken upstream commands/links.
3. Marketing site + feedback routing — needs Kata release URLs, store listings, and legal entity decisions before legal pages.
4. Provider/MCP identity — lockstep commit across adapters + tests (Remaining T3 user-facing branding #118 calls this out explicitly).
5. Telemetry + themes — telemetry is low user impact; themes need a prefs migration decision first.

No PRs reference this issue yet.

[gannonh]

Hygiene

Unparented from closed Phase 1 #109 so this issue can own children without nesting two levels.

Rewrote the punch list into an epic. First child: #127 (update copy, pairing labels, install/docs). Remaining slices stay deferred until a product call or a later child spec.

Internal identifiers stay on #119.

[gannonh]

Follow-up after adversarial review of child #127:

• Parent stays a deferred epic (status:draft). Do not approve or build this issue.
• Verification on this parent now defers to each child's name table + grep. Do not treat Remaining T3 user-facing branding #118 as greppable scope.
• Constraints now name T3_* env vars as Retain upstream-compatible internal T3 identifiers #119, so a copy child cannot "fix" scripts/dev-runner.ts identifiers.
• Later slices called out as independent user outcomes with their own blockers, not a layer waterfall.

Slice 1 to approve (or revise) is #127 only.

[gannonh]

#119 routing

#119 retains upstream-compatible internal and durable T3 identifiers. The following user-visible items move to this epic for future approved work:

• The t3-connect profile route or label when it appears in shipped navigation.
• The ~/.codex-t3/personal Settings placeholder.
• User-facing T3/Codex wording in docs/user/providers-codex.md and shipped T3 labels or user-facing examples in current Connect documentation. Contributor-only internal terminology stays with Retain upstream-compatible internal T3 identifiers #119.
• t3.chat when it appears in agent tool text or other shipped examples. Reserved test domains stay with Retain upstream-compatible internal T3 identifiers #119.

#119 retains t3.json, wire values, checkpoint refs, browser storage keys, internal T3_* variables, binary names, logs, CSS identifiers, temporary names, and fixtures. It does not change runtime code in this slice.