fix(spec): review-pass fixes across the truth stack - #961
Open
leeovery wants to merge 1 commit into
Open
Conversation
Findings from the eight-dimension review of stack #956, verified and applied together: Routing (process-review-findings.md): the route lane gains the two-way exit its siblings carry, B gains the all-settled branch, and the auto flip re-enters B per finding — a routed or gap finding can no longer be swallowed by the bulk sweep, and a tracking file whose last finding routed now completes. The on-return step lands by what actually happened: a landed resolution re-aligns the spec (announced, the approval having preceded it at the stop), a record-settled point aligns to the governing decision, a held-doc queue leaves the spec copy alone — the reopened source's stale row carries the reconcile. Incoherence reference: caller-neutral prose (two callers now), the four-move count, a complete dichotomy in the measurement branch (nothing-leans-on-it vs Otherwise), and a landing idiom for decisions the document never made (new subtopic section — no timeline to revise). Agents: gap analysis routes open-decision markers as Unsourced decision instead of landing them spec-side; both older agents' one-concern fences name the claims lane; the claims opener drops its rationale. Consistency: three-phase wording lands in the two remaining stale asserts, specification-format's completion criteria, the shared tracking format's field guidance, commands.md (spec-review-gate entry, finding category contract, gate-force qualifier), CLAUDE.md's phase pointers, and four docs passages. The measured-claim shape is ruled command-alone-in-span (re-runs by copy) everywhere it is stated. Engine: FINDING_CATEGORIES hoisted with the routed-category refusal named, a/auto suppressed on a category-forced gate, the banner rule restored, and spec-review A drains in-progress tracking before opening a cycle. Review-time measurement gets a read-only Bash set on the three skills' allowed-tools (investigation precedent). Menu adoption (per review ruling): spec-completion's two consent gates and research document-review's carry-note gate move to engine surfaces (render spec-completion-gate --variant assessment|signoff, render carry-note-gate), with tests, catalogue entries, and a simulation extension including the refusal the old comment claimed. Coverage: two new prose cases — spec-routes-an-unsourced-decision (the 16-undiscussed-decisions lane end to end, landing as a new subtopic) and research-review-corrects-a-false-claim (the research claims sweep) — with their stubs; the three earlier recipes drop their inert git calls; the research document-review ratchet pin shrinks to zero. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
The verified findings from the eight-dimension review of stack #956, applied together, plus the three rulings from the pass:
Routing —
process-review-findings.md's route lane gains the two-way exit its siblings carry (a tracking file whose last finding routed now completes), B gains the all-settled re-entry branch, and theautoflip re-enters B per finding so routed and gap findings can never ride the bulk sweep. The on-return step lands by what actually happened: landed resolution → spec re-aligns (announced — approval preceded it at the stop); record-settled → align to the governing decision; held-doc queue → spec copy untouched, the reopened source's stale row carries the reconcile.Incoherence reference — caller-neutral for its two callers, the four-move count, a complete dichotomy in the measurement branch, and a landing idiom for decisions the document never made (new subtopic section — no timeline to revise).
Agents — gap analysis routes open-decision markers as
Unsourced decisioninstead of landing them spec-side; the one-concern fences name the claims lane; the claims opener drops its rationale.Consistency — three-phase wording in the two remaining stale asserts,
specification-format's completion criteria, the shared tracking format,commands.md(new surface entries, thecategorycontract, the gate-force qualifier), CLAUDE.md's pointers, four docs passages. Measured-claim shape ruled command-alone-in-span everywhere.Engine —
FINDING_CATEGORIEShoisted with the routed-category refusal named,a/autosuppressed on a category-forced gate, the banner rule restored, spec-review drains in-progress tracking before opening a cycle.Rulings applied (per review): menu adoption —
spec-completion-gate(assessment/signoff) andcarry-note-gatesurfaces replace the three prose menus, with tests, catalogue entries, and the simulation refusal; read-only measurement Bash on the three skills' allowed-tools; two new prose cases —spec-routes-an-unsourced-decisionandresearch-review-corrects-a-false-claim— with stubs; the earlier recipes drop their inert git calls.Gates:
npm test2289 pass,test:cligreen, typecheck clean, conventions ratchet tightened (research document-review pin → 0).Stacks on #960. Part 7 of the idea-40 stack.
🤖 Generated with Claude Code