Close out Lane G status records (#145)

[mchwang]

Summary

Closes #145. Lane G's four steps (G1–G4) merged as PRs #148, #150, #151 and #152, and docs/architecture.md already marks lane G done. Several documents still described the Planning screen as unbuilt, gave lanes G and H active ownership of files they had released, or described Plans screen features that were never built. F5 (#149) integrated after G4 and kept "the Planning … screens" under "Still deferred".

This PR changes documentation only. It makes no code, test or configuration changes.

• README: lists the Plans screen as built (plan import, plus agent drafts and suggestions with revision-bound Apply). Drafts and suggestions need a configured github review, not the demo (web/planning.ts, web/cli.ts). Ask, planning and the runner are named as the Claude Code container users.
• Design doc (docs/designs/codeboost-plan-indexed-review.md):
  • Summary: moves the Plans screen to "Built", with "Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155 lists unbuilt parts". The header and the 2026-09-26 "Lane status" table now point to docs/architecture.md for current lane state.
  • Build step 6: "Done for T18", with the unbuilt parts listed (Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155). "What to do next" adds E4 and the Plans screen.
  • "When the plan is wrong": a note says the runner does not yet save a plan-change proposal and the screen cannot show one (Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155).
  • Parallel build lanes: a "Current state" note says the lane plans record the original schedule and that B0, C, D, E, G, H and K are done. The Wave 2 G row is marked Done. Integration rule 2 now describes the wave 1 and wave 2 ownership as past, and points to "Task resumption and ownership" in AGENTS.md.
  • Design review and DT status: the review note says Plans and Issues were built without that review. The DT status list removes Plans from the plain-text menu links and records the mockup differences.
• docs/plan-format.md: its contract-mirroring text is unchanged; it must match the immutable v1 contract in schema/versions/1/semantics.md. A new section, "What the shipped Plans screen does today", lists the differences:
  • features not built yet (Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155);
  • suggestion cards that do not yet conform to the v1 contract (Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159);
  • the real control names.
• docs/implementation/planning-screen.md: adds a Handoff section listing the files lane G releases, the files that stay with lane F, and the open follow-ups (Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155, Revisit E3's early 'cancelled' write now that G4 wired the planning endpoints #156, Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159), and records that lane F fixed the import size gap (Align plan import request limit with valid schema payloads #146) in PR Size the plan import request limit from the parser bound (#146) #158. It also describes the shipped sibling-card behaviour and cites Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159.
• docs/implementation/issue-prioritization.md: lane H's ownership text is now in the past tense, and H4a's and H4b's original plans are stated as plans.
• docs/implementation/runner-lifecycle.md: decision 2's "revisit at G4" now links Revisit E3's early 'cancelled' write now that G4 wired the planning endpoints #156.

Follow-up issues opened during review:

• Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155: Plans screen design features that are not built (paste import, in-screen editing, amendment proposals, warnings, answering questions, whole-plan approval).
• Revisit E3's early 'cancelled' write now that G4 wired the planning endpoints #156: revisit E3's early cancelled write now that G4 has wired the planning endpoints.
• Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159: the shipped Plans screen does not conform to the v1 suggestion-card contract. Sibling cards lose Apply instead of showing it disabled, and there is no Dismiss on each card. The fix is either restoring conformance or adding a new semantic version.

Validation

Exact base/head: a764e1983ddb49957fdb1092e801afe8a4fcdeb7 → befe2048e7f51b1b386c63c394d18608b72549fe (validated head). Main moved during review: PR #158 (fix for #146) landed, and the branch merged it. Earlier pairs were on base ae9fc3d and are intermediate checkpoints: Copilot rounds 1–4 reviewed 5204326, 7e7abd3, 8e26937 and 82b7e5c. The round 5 request on d30cd47 was superseded by the base change.

• git diff --check: passed
• Every factual claim was checked against GitHub (gh pr view, gh issue view, PR file lists) and the code (web/public/index.html, web/public/app.js, web/planning.ts, web/cli.ts, web/server.ts:593, runner/execution.ts, runner/store.ts).
• Tests, typecheck and real-Docker were not run, because the change touches no code. CI runs the full suite.

Review record

Every independent round was a fresh general-purpose agent reviewing the full diff from ae9fc3d. From round 12 on, reviewers were limited to read-only git commands.

• Author self-review: a full-diff pass before the first independent round.

• Rounds 1–10: stale and contradictory status lines, the released file list, Align plan import request limit with valid schema payloads #146 ownership wording, the Wave 2 G row, README demo limits. All fixed. Declined: line 569 "the runner (lane F) is next" belongs to lane F's status, which the open F6 PRs (F6a: push the rewritten head behind a durable marker #153) own. Also declined: the "Planning screen" (build step) vs "Plans screen" (UI label) naming split; round 2 asked for the opposite change, and line 570 links the two names.

• Round 11 (..5204326): clean. It authorized Copilot round 1.

• Copilot round 1 (5204326): 1 low finding. Wave 2 guidance and rule 2 still read as G's active ownership of the web files. Fixed. The earlier review briefs had wrongly put those lines out of scope.

• Rounds 12–17: the ownership history I added was imprecise:

  • who held web/server.ts;
  • G3's server change;
  • web/cli.ts;
  • H4b editing web files until just before G1.

  Fixed by naming G3's change and keeping only facts the history supports.

• Round 18 (..7e7abd3): clean. It authorized Copilot round 2.

• Copilot round 2 (7e7abd3): 0 open findings; round 1's thread resolved. One "previously missed" item: rule 2 still stated C's and D's wave 1 assignments as active. Fixed by stating them as the wave 1 plan.

• Rounds 19–33:

  • Older lane rows read as current. Fixed with the "Current state" note, including B0.
  • Plans screen features were described as present but not built (paste, editing, proposal view, warnings, questions, approval). Marked, and Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155 opened.
  • The unrecorded decision 2 revisit. Revisit E3's early 'cancelled' write now that G4 wired the planning endpoints #156 opened.
  • Next steps, the DT status, the design-review note and the Settings wording.
  • Exact control labels and layout.
  • Lane H's tense, and an H4b plan wrongly turned into a past-tense claim.

  All fixed. Declined (round 26): adding the drawer vs side-panel layout difference to Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155, because it is a layout choice already recorded in the DT status list, not a missing feature. From round 26, reviewers reported target-design gaps only when the text presents a feature as available now; Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155 tracks the rest.

• Round 34 (..8e26937): clean. It authorized Copilot round 3.

• Copilot round 3 (8e26937): 1 finding, valid. My round-24 edit rewrote plan-format.md's sibling-card wording to match the shipped screen, which contradicts the immutable v1 contract (schema/versions/1/semantics.md). The guide may not override that contract. The same was true of my other wording edits in that guide. Fixed by restoring the guide's contract text to the base version, adding a separate "What the shipped Plans screen does today" section, and opening Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159 for the non-conformance. The shipped regression asserts that sibling Apply is absent (toHaveCount(0)), not disabled.

• Round 35 (..eded1eb): contract check passed, plus two optional P3s: cite Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159 in the summary and "What to do next", and list the Align plan import request limit with valid schema payloads #146 limit. Both applied.

• Round 36 (..82b7e5c): clean. It authorized Copilot round 4.

• Copilot round 4 (82b7e5c): round 3's thread resolved; 1 new finding. The shipped-differences list in plan-format.md left out the amendment-proposal flow that Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155 tracks. Fixed.

• Round 37 (ae9fc3d..d30cd47): clean. It authorized the Copilot round 5 request, which the base change then superseded.

• Base change: main moved to a764e19 (PR Size the plan import request limit from the parser bound (#146) #158 fixed Align plan import request limit with valid schema payloads #146). The branch merged it cleanly. The diff's statements that Align plan import request limit with valid schema payloads #146 was open and that imports were limited to 16 KiB became wrong. Fixed: the handoff records the fix, and the shipped-differences bullet was removed.

• Round 38 (a764e19..befe204): diff correct, merge preserved Size the plan import request limit from the parser bound (#146) #158. One P3 on the issues: the Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155 and Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159 acceptance sections did not name every doc location that cites them. Fixed in the issue bodies; the diff did not change.

• Round 39 (a764e19..befe204): clean, no findings. This pass authorizes Copilot round 6 on a764e19..befe204.

Review-lesson audit

• Stale and contradictory status (rounds 1, 4–6, 10, 19, 22, 28, 30; Copilot 1 and 2): covered by the External status confirmation rule. Requested, built, validated and complete are distinct states.
• Unverified claims, including history inferred from plans (rounds 2, 3, 7–9, 13, 15–17, 21, 23–25, 27, 29, 31, 33): covered by the existing rule to treat every behavioural claim in docs as something to verify.
• The F5 integration that kept the Planning screen under "deferred": covered by the Stacked pull requests rule to preserve both sides' documentation.
• Deferred work found in review (Plans screen: design features not built (paste import, editing, proposals, warnings, questions, approval) #155, Revisit E3's early 'cancelled' write now that G4 wired the planning endpoints #156, Plans screen does not conform to v1 suggestion-card contract (sibling Apply, per-card Dismiss) #159): covered by "Record newly discovered deferred work as issues before closing the implementing task".
• Review briefs that excluded lines Copilot then flagged: covered by the rule that each pass examines the full diff and every claim. Excluding them was a process slip, not a missing rule.
• A review agent ran git checkout in the author's worktree, so one push sent an older head. Caught by comparing the PR head with the reviewed head. Covered by "Run final validation against the exact pushed head". Later briefs forbid commands that change the worktree.
• Editing a guide that mirrors a released contract (Copilot 3): covered by the existing rule in docs/plan-format.md that the guide cannot override a released snapshot, and by the rule to verify every behavioural claim. The edit was made without checking semantics.md, which was a process slip.
• No new AGENTS.md/CLAUDE.md rule is needed.

🤖 Generated with Claude Code

[Copilot]

🟡 Changes recommended

The design document still contains conflicting active guidance that G owns and may continue work on files it now says were released.

1 open finding

• Resolve conflicting historical and current web ownership guidance · New

What changed in this PR

Updates status and handoff documentation to reflect Lane G’s completed Plans screen delivery.

Changes:

• Marks the Plans screen and build step 6 as complete.
• Documents Lane G’s released files and remaining issue #146.
• Clarifies historical versus current lane status records.

File	Description
README.md	Adds the Plans screen and its configuration requirements.
docs/​implementation/​planning-screen.md	Records Lane G’s handoff and remaining import-limit issue.
docs/​designs/​codeboost-plan-indexed-review.md	Updates Lane G completion status and ownership history.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🔵 Needs a closer look

The design document still presents completed wave-1 ownership assignments as active guidance.

0 open findings

1 resolved since last review

• Resolve conflicting historical and current web ownership guidance

Previously missed (1)

In code that hasn't changed since last review

Use past tense for completed wave-1 ownership assignments

docs/​designs/​codeboost-plan-indexed-review.md:1970

The paragraph still presents two completed wave-1 assignments as active rules: C “owns” common configuration and D “exclusively owns” the isolation workflow. Both lanes are already marked done in docs/architecture.md:410-413, and F1 has merged, so this can still send future editors to obsolete owners despite the new current-ownership guidance earlier in the same line. Put these remaining historical assignments in past tense as well.

🧠 Review effort: Balanced

[Copilot]

🟡 Changes recommended

The documented sibling-card behavior conflicts with the immutable normative v1 contract.

1 open finding

• Resolve conflict between v1 Apply-control contract and shipped behavior · New

🧠 Review effort: Balanced

[Copilot]

🟡 Changes recommended

The shipped-screen gap list omits the unbuilt amendment-proposal flow tracked by #155.

1 open finding

• Document missing amendment-proposal flow in shipped-screen differences · New

1 resolved since last review

• Resolve conflict between v1 Apply-control contract and shipped behavior

🧠 Review effort: Balanced

[Copilot]

🟢 Approval recommended

The documentation changes consistently match the implementation, architecture status, and tracked follow-up issues.

0 open findings

1 resolved since last review

• Document missing amendment-proposal flow in shipped-screen differences

🧠 Review effort: Balanced

[Copilot]

🟢 Approval recommended

The documentation matches the verified implementation and clearly tracks remaining gaps.

0 open findings

🧠 Review effort: Balanced