docs(sidechat): record the followUpIsPlan exclusion with a pinning test (#6754)

[iamwhatever]

Problem / Motivation

deriveFollowUpOptions (website/src/app-sdk/protocol/options.ts) returns followUpOptions, followUpIsPlan, and followUpSourceKey. website/src/pages/chat/SideChat.tsx destructures only followUpOptions, silently dropping followUpIsPlan. The drop is correct in effect — the side panel is not a plan-capable host — but nothing records that: no comment states the rationale, and no test pins the behaviour. A contributor auditing plan-chip coverage finds an undocumented drop and either re-derives the rationale from scratch or "fixes" it by wiring dispatch into a host that has no mode source to gate it.

Why it matters

The undocumented exclusion is a trap for future contributors. Wiring plan dispatch into SideChat without a mode gate would let any plan-shaped side answer cancel or advance the parent slot's real orchestrator plan — a live cross-conversation mis-dispatch. The record (comment + pinning test) is what turns "silent drop" into "reviewed, deliberate exclusion" and makes any future wiring attempt fail a test instead of shipping.

What changed (motivation → approach → change)

Applies the same-size record pattern as PR #6748 does for ChatEmbed (sibling issue #6057 covers the same drop there). Two deliverables, zero runtime behaviour change:

1. A block comment at SideChat's deriveFollowUpOptions destructure recording the load-bearing facts: a side turn runs as an aside to the parent session and is never an orchestrator turn, so useComposerDraft owning a plan-shaped chip (pick → edits the draft, amendable before send) is correct behaviour, not a fallback; the dispatch path gates on the host slot's mode (ChatPage reads effectiveMode === 'orchestrator' off the slot record), which this panel does not read today; and an unconditional dispatch would let a plan-shaped side answer cancel or advance the parent's real plan.
2. A pinning test (website/src/test/SideChat.planExclusion.test.tsx), mirroring the ChatEmbed plan-exclusion test shape, with a shared SIDE_MESSAGES fixture feeding both the rendered store and the premise derivation so they cannot drift apart.

Deliberately out of scope (same as #6748): wiring usePlanActionMutation into SideChat. Path B parity is a product decision, not this change.

Tests

New file website/src/test/SideChat.planExclusion.test.tsx, three assertions in one scenario:

• Premise pin: deriveFollowUpOptions(asTranscript, false).followUpIsPlan === true, where asTranscript is mapped from the same SIDE_MESSAGES array the store renders, through the same role/cls expression SideChat's transcript memo applies (non-streaming, non-error case). Without this the test would pass vacuously the moment the fixture stops being plan-shaped.
• Composer-draft path: clicking the Go chip lands the label in the side composer's draft (input.value === 'Go').
• No dispatch: api.planAction is never called (and api.sideTurn is not called either — picking is an edit, not a submit).

Red-before proven: a plan dispatch was temporarily wired into SideChat's chip handler (if (followUpIsPlan) { void api.planAction(slot, o); return }); the test then failed (expected '' to be 'Go' — the draft assertion reds because dispatch short-circuits the draft edit). The scaffolding was reverted; only the comment and the test remain in the diff.

Full frontend suite on the rebased branch: 1638 test files, 25,938 passed, 0 failures (npx vitest run). npx tsc -b clean.

Manual verification

N/A — unit coverage sufficient: the change is a comment plus a test; the pinned interaction (chip click → draft edit, no dispatch) is exactly what the test drives through the real component.

Screenshots / video

Why no screenshot: the diff adds only a code comment and a test file; no rendered output changes anywhere.

N/A — this change adds a comment and a test and alters no rendered output. No pixel changes anywhere; explicitly not a skipped screenshot gate.

Related Issues

Fixes #6754

Checklist

• At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|refactor|perf|test|chore|ci|build|revert: ...)
• Existing tests pass and new tests added for new functionality
• Self-review completed; code follows project style guidelines
• Documentation updated (if applicable)
• No secrets, credentials, or internal references in the diff

[github-actions]

UX Review (Fable 5) — ✅ PASS

UX-level review of 7d7596f5a657973ac9b349cb8797db56a69e7fc5 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: PASS

No user-facing change: the diff adds a code comment and a pinning test; no string, pixel, or interaction a user experiences is altered.

[UX-REVIEWED] 7d7596f

[github-actions]

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 7d7596f5a657973ac9b349cb8797db56a69e7fc5 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 7d7596f

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 7d7596f5a657973ac9b349cb8797db56a69e7fc5: <one-sentence reason>

[github-actions]

Design Review (Fable 5) — ✅ PASS

Design-level review of 7d7596f5a657973ac9b349cb8797db56a69e7fc5 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

A real contributor trap, closed with the cheapest sound mechanism: a rationale comment plus a non-vacuous pinning test, zero runtime change.

The test's design holds up: the premise pin (followUpIsPlan === true derived from the same fixture the store renders) prevents the vacuous-pass failure mode that usually sinks absence-pinning tests, and the draft assertion reds only on an unconditional dispatch wiring — a future properly mode-gated wiring would still pass, which matches the stated scope (Path B parity stays a product decision). The comment's claims check out against ChatPage.tsx:7812 (followUpIsPlan && … && effectiveMode === 'orchestrator'). Pattern is consistent with the ChatEmbed sibling. No design-level concerns.

[DESIGN-REVIEWED] 7d7596f

[github-actions]

First Principles Review (Fable 5) — ✅ PASS

Premise-level review of 7d7596f5a657973ac9b349cb8797db56a69e7fc5 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

All claims verified. The key fact grounding this change: usePlanActionMutation.ts:146 records defect #5893 — ChatPane also dropped followUpIsPlan and there it was a real bug, fixed by wiring dispatch. That prior incident is exactly what would mislead a contributor into "fixing" SideChat the same way, so the harm is derived from a reported defect, not symmetry. Counts: 4 hosts call deriveFollowUpOptions (ChatPage, ChatPane consume followUpIsPlan; ChatEmbed, SideChat drop it); the ChatEmbed sibling drop is declared and tracked (issue #6057 / PR #6748); no existing test pins SideChat's exclusion (7 prior SideChat test files, none touch planAction); the diff ships zero runtime change and zero new public surface.

First-Principles-Verdict: PASS

Both items pin a real hazard with a named precedent — #5893 shows this exact drop was once a live bug in a sibling host — at zero runtime cost.

What this change ships

Intent: make SideChat's deliberate drop of followUpIsPlan un-mistakable for the bug it resembled in ChatPane (#5893). This is an ADDITION (a record: comment + test), declared as such.

1. Comment at the destructure explaining why the drop is deliberate — justified (derived from defect ChatPane renders plan-approval follow-up chips but cannot dispatch plan actions #5893, usePlanActionMutation.ts:146)
2. New test failing if plan dispatch is ever wired into SideChat — justified; no prior test covers it (grepped planAction across 7 SideChat test files: 0 hits)

Both items are declared; no config, flag, export, or behavior rides along (diff verified: SideChat.tsx hunk is comment-only). The one sibling of the root cause — ChatEmbed's identical drop — is counted (1) and explicitly tracked in #6057/#6748, so the point-patch scope is declared and deferred, not silent.

[FIRST-PRINCIPLES-REVIEWED] 7d7596f

[github-actions]

Opus 4.8 Review — ✅ no blocking findings

Reviewed 7d7596f5a657973ac9b349cb8797db56a69e7fc5 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 7d7596f

Verdict parsed from the review's SHA-scoped output markers for commit 7d7596f5a657973ac9b349cb8797db56a69e7fc5.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 7d7596f5a657973ac9b349cb8797db56a69e7fc5: <one-sentence reason>

[bolichen97]

Approved after a full-diff review (scope match, no out-of-scope files, security surface checked, tests verified non-vacuous). Review notes available on request.