Console: ENGINE_META omits little-coder, so its panes render the raw id

[dichovsky]

web/view-model.ts:141-161 defines ENGINE_META for seven of the eight Participant CLIs — claude-code, codex-cli, gemini-cli, copilot-cli, antigravity-cli, pi-cli, opencode-cli. little-coder is absent, even though src/participants.ts:7-16 and PARTICIPANT_TARGETS (src/platforms/registry.ts:20-29) both declare it.

The consequence is at web/view-model.ts:170: engineMeta() falls through to the unknown-engine branch, so an Agent joined as little-coder is rendered in the Console with the neutral fallback badge and the raw id string rather than a display name and tint, exactly as if it were an unrecognized platform.

Found while reviewing #56 (which resolved #26). It was deliberately left out of that change: #56 is docs-and-metadata only, this is a code defect, and .github/ISSUE_TRIAGE.md's scope rule keeps a distinct problem out of an unrelated diff.

Impact: the Console silently presents a fully supported Participant as unrecognized. This is the same roster-drift class as #26 but on the code side, so it is invisible to the documentation sweep that caught the rest — and tests/unit/docs-facts.test.ts does not cover it either, since that guard compares docs-site/generated/facts.json against PARTICIPANT_TARGETS and never reads web/.

Acceptance: ENGINE_META covers all eight ids, with a little-coder entry whose display name matches the one used elsewhere in the repo ("Little Coder", per docs/design/setup-integration.md). Ideally the gap cannot silently reopen: a test asserting that ENGINE_META's keys equal PARTICIPANT_IDS would fail today and would catch the next engine landing in the registry without a Console badge. Note the display names in ENGINE_META currently diverge from the registry's ("Codex", "Gemini", "Copilot" vs the registry's fuller names) — whether to reconcile those too, or only to close the missing-key gap, is a judgment call for the implementer.

[dichovsky]

Triage: re-verified against main @ d311767. Confirmed as-is — both halves reproduce exactly as filed.

• ENGINE_META (web/view-model.ts:141-161) has seven keys: claude-code, codex-cli, gemini-cli, copilot-cli, antigravity-cli, pi-cli, opencode-cli. little-coder is absent.
• The registry declares eight: PARTICIPANT_IDS (src/participants.ts:7-16) and PARTICIPANT_TARGETS (src/platforms/registry.ts) both include little-coder, backed by src/platforms/little-coder.ts.
• The consequence is at web/view-model.ts:169-178: engineMeta() guards with platformId in ENGINE_META, so little-coder falls to the unknown branch and renders label: 'little-coder' with the neutral · glyph and #5b6675/#eef0f3 tint — identical to an unrecognized platform.
• The gap is genuinely unguarded: grep -rn ENGINE_META across src/, web/, tests/, and docs-site/ returns only the definition (:141) and the one read (:170). Nothing asserts key coverage.

One delta against the body (noted here, body left as written): the "Ideally the gap cannot silently reopen" test cannot compare ENGINE_META's keys to PARTICIPANT_IDS as written, because ENGINE_META is module-private — it is const, not export const (web/view-model.ts:141). A guard needs either an export or an engineMeta(id).label !== id assertion looped over PARTICIPANT_IDS, which stays within the current surface.

Labels applied:

• bug — the Console presents a Participant the registry declares supported as unrecognized. Under the triage table this is code behavior contradicting the authoritative registry, not documentation drift; the docs side of the same roster class is Participant roster drift: README, CLAUDE.md/AGENTS.md, SRS, product-spec, and npm metadata disagree with the eight-CLI registry #26 (PR docs: align every roster introduction on the eight Participant CLIs #56) and is not affected by this.
• good first issue — the required fix is one entry in one object literal in a single file, with the display name already fixed by precedent ("Little Coder", docs/design/setup-integration.md). The reconciliation of the other labels against the registry's fuller names is explicitly optional in the Acceptance and should be left out rather than folded in.

Not a duplicate: gh issue list --state all --search on the roster terms returns #26, #58, and this one. #26 and #58 are prose restatements of the roster in documentation and a test docstring; this is the only one where code behaves differently as a result.

---

Generated by Claude Code