boocode/docs/plans/post-review-backlog/artifacts/implementation-iteration-history.md

# Implementation Iteration History: Post-Review Backlog (F1–F9)

<!--
This file records how the implementation plan for the post-review backlog evolved across discussion rounds.
Committed decisions live in [implementation-decision-log.md](implementation-decision-log.md) and the primary
plan lives in [../feature-implementation-plan.md](../feature-implementation-plan.md). It also consolidates
the project-manager's per-round facilitation output (claim ledger, Open Questions, spec-maturity tags).
The loop converged in one round; remaining unknowns are spec-level, not resolvable by more specialist rounds.
-->

## R1: Parallel six-specialist backlog review

- **Specialists engaged:** on-call-engineer, behavioral-analyst, software-architect, test-engineer, user-experience-designer, junior-developer, project-manager (coordinator). Team size: large (cross-subsystem; user chose scope "everything we discussed"). Round cap 3; converged in 1.
- **New input provided:** Initial inputs — the [scope brief](../scope-brief.md) (ground-truth spec stand-in; two items live-verified 2026-06-02 with file:line evidence) and the [discovery notes](.discovery-notes.md) (per-item code touch points). No prior round; this is the initial sweep.
- **Claim ledger:** (consolidated, deduped — see [synthesis-input.md](synthesis-input.md) for the full table)

  | # | Claim | State | Spec-maturity |
  |---|-------|-------|---------------|
  | C1 | F1 cancel route never aborts external child; no registry/export | Evidenced | plan-level |
  | C2 | F1 catch blocks leave message `streaming`; success path writes `complete` on abort — fix same batch | Evidenced | plan-level |
  | C3 | F2 = prune-now-minimal: unexport 8 zero-caller symbols, keep extractToolCallBlocks+stripToolMarkup | Evidenced | plan-level |
  | C4 | F2 `<invoke>`-text fallback untested → add gate test before prune | Evidenced | plan-level |
  | C5 | F3 optional logger param, do with F2 (same file) | Evidenced | plan-level |
  | C6 | F6 stall-timeout via AbortSignal.any, 90s; NO retry (non-idempotent deltas) | Evidenced | plan-level |
  | C7 | F7 inline MCP tool, messages_with_parts, role!='system', limit 50/200 | Evidenced | plan-level |
  | C8 | F9 delete SPA block, keep routes; GET / unprobed → safe | Evidenced | plan-level |
  | C9 | F4 hook-firing in unattended mode UNVERIFIED; goose hook mechanism unknown | Anecdotal (premise) | spec-level |
  | C10 | F4 dedup rule: confirm running before `blocked`; suppress hook `done` | Evidenced | plan-level |
  | C11 | F5 pinned @opencode-ai/sdk exposes no compaction arm → blocked on capability check | Evidenced | spec-level |
  | C12 | F5 UI treatment sentinel-row vs ephemeral-frame | Disputed | spec-level |
  | C13 | F8 no line-selection infra; diff source ambiguous; needs own spec | Evidenced | spec-level |

- **Open Questions raised:**
  - F1: terminal state (→ D-7, `cancelled`); registry key (→ D-1, `taskId`); shared finalize helper (→ D-1, yes); warm re-throw on abort (→ D-1, short-circuit on `ac.signal.aborted`).
  - OQ-F2a (sidecar jinja) → moot under D-3 option A; OQ-F2c (a vs b) → D-3 option A.
  - OQ-F6a/b/c → D-5 (AbortSignal.any, no retry, 90s).
  - OQ-F7a (session vs chat id) → D-6 (both, chat_id optional, + limit).
  - OQ-F9a (GET / probe) → D-11 (unprobed, safe).
  - OQ-F4a (hooks fire unattended?), OQ-F4b (goose hook format) → UNRESOLVED, spec-level → OI-1, route F4 to plan-a-feature (D-12).
  - OQ-F5a (SDK compaction event existence/name) → UNRESOLVED, capability check → OI-2, blocks F5 (D-12).
  - OQ-F5b (sentinel vs ephemeral UI) → UNRESOLVED → OI-3, settle once event confirmed (D-12).
  - OQ-F8a/b/c (diff source, serialization, new viewer) → UNRESOLVED, spec-level → OI-4, route F8 to plan-a-feature (D-12).
  - OI-5 (F1 best-effort session-stop leg) → non-blocking, decided at implementation.
- **Spec-maturity tags:** plan-level — C1-C8, C10 (9 claims). spec-level — C9, C11, C12, C13, plus OQ-F4b and OQ-F8a (≥5 across junior, behavioral, UX). **Spec-maturity gate TRIPPED**, concentrated in the three WANT items F4/F5/F8; F1/F2/F3/F6/F7/F9 are all plan-level and ready. No T#-contradictions.
- **Resolution source:** evidence (Step 6 specialist findings) for every plan-level OQ (F1, F2, F3, F6, F7, F9); user input for the gate disposition (Sam's standing override → D-12); deferred-to-spec for OQ-F4a/F4b, OQ-F5a/F5b, OQ-F8a/b/c (recorded as OI-1..OI-4, routed out rather than resolved in this loop). The YAGNI gate ran during synthesis: F6 retry, F2 option B, F4 interface, F5 extra compaction arms, F7 reader interface all deferred.
- **Decisions produced:** D-1, D-2, D-3, D-4, D-5, D-6, D-7, D-8, D-9, D-10, D-11, D-12 (all 12; the loop converged in one round so every decision originates here).
- **Changed in plan:** all sections (initial authoring) — Source Specification, Outcome, Context, Implementation Approach (TIER 1 READY / TIER 2 BLOCKED), Decomposition and Sequencing, RAID Log, Testing Strategy, Security Posture, Operational Readiness, On-Call Resilience Posture, Definition of Done, Specialist Handoffs, Deferred (YAGNI), Open Items, Summary.
- **Project-manager next-step recommendation:** Go to synthesis (done — this plan). Build the READY cluster in order F1 → F2+F3 → F6 → F7 → F9 as sequential patch tags; route F4 and F8 to `plan-a-feature` and F5 to an `@opencode-ai/sdk` capability check before any build on those three.