[bompus]

Two gaps found by the bot review of the setupLinuxHost adoption of the unbuilt-stage rule (#637).

Per-offer status. A stage line counted an offer as unanswered only when asked|pending|unanswered matched anywhere in the line, trigger included. So Approved, not built: export; Offer NEXT-1 sent; awaiting user reply; trigger: user reply gave no unanswered offer, and three stages approved in separate offers with trigger: pending local build gave three and a false walk-through warning. Each offer is now judged by its own clause (split at ; and ,; a clause naming no offer, such as awaiting user reply, belongs to the offer before it), and the trigger: clause is not read.

Recheck warning. rechecks() warned about an undated wait on another thread only for a deferred [~] line or a sent-question line, but a stage is an open [ ] line. An open stage whose trigger: names an upstream or other pull request, a maintainer or another session now gets the same missing-recheck: warning, and a valid recheck: YYYY-MM-DD clears it.

Tests cover both examples, a line naming one approved and one unanswered offer, and the recheck cases; both new tests fail on the old script. bun run check passes; bun test test/ (1330 tests, 19 skipped) and bun test skills/ (1009, 11 skipped) pass with 0 failures.

The third finding from that review, the OpenCode scope lacking the approved-stage convention, lives in setupLinuxHost's overlay (composeOpencodeScope, SPANS in overlay/house-rules/generate.ts), not in this repository.

[bompus-ai-review]

Bompus AI Review (GPT-6.1 Terra, max reasoning): 6 finding(s).

[bompus-ai-review]

Bompus AI Review could not run on f4bca77: no reviewer seat finished the review (GPT-6.1 Terra: usage limit reached, 5-hour 100% until 11-14 00:41 UTC; GPT-6.1 Sol: usage limit reached, 5-hour 100% until 11-14 00:41 UTC; Muse Wonder Full: usage limit reached, Weekly 99% until 11-16 00:00 UTC; Gemini 4 Argon: usage limit reached, Gemini 4 Argon Weekly 100% until 11-14 23:30 UTC; SWE-2 High: usage limit reached, rate limit until 23:02 UTC; second seat, Gemini 3.8 Flash: usage limit reached, Gemini Weekly 100% until 11-14 01:41 UTC; Muse Wonder Full: usage limit reached, Weekly 99% until 11-16 00:00 UTC; Gemini 4 Argon: usage limit reached, Gemini 4 Argon Weekly 100% until 11-14 23:30 UTC; SWE-2 High: usage limit reached, rate limit until 23:02 UTC).

Seats that may review this pull request, in order: GPT-6.1 Terra > GPT-6.1 Sol > Muse Wonder Full > Gemini 4 Argon > SWE-2 High. If none of those finishes, a second seat may review alone: Gemini 3.8 Flash > Muse Wonder Full > Gemini 4 Argon > SWE-2 High.

Nothing was reviewed. The service tries this head again by itself after about 23:02 UTC; a new commit meets the same seats, so do not push one to retry. Merging before then needs the owner's exception.

[bompus-ai-review]

Bompus AI Review (Gemini 3.8 Flash, high reasoning): 0 finding(s).

Second reviewer only: every first reviewer failed on this commit, so this is Gemini 3.8 Flash's review alone.

Review 2: findings below high are posted only within 3 lines of a change since 6c4d6e4.