This is an automated email from the ASF dual-hosted git repository. spmallette pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/tinkerpop.git
commit 0192d91ee298ba2634fb50e8b1276393e0104631 Author: Stephen Mallette <[email protected]> AuthorDate: Fri Oct 9 15:58:58 2026 -0400 Have tinker-review read beads critically Recorded decisions are now claims the reviewer verifies against the code, changed code no decision covers gets a closer read, and design trade-offs settled only in the PR thread are raised as open questions. Assisted-by: Claude Code:claude-opus-5-5 --- .skills/tinker-review/playbooks/general.md | 29 +++++++++++++++++++++++++---- 1 file changed, 25 insertions(+), 4 deletions(-) diff --git a/.skills/tinker-review/playbooks/general.md b/.skills/tinker-review/playbooks/general.md index 9d8cae833e..2387c8a081 100644 --- a/.skills/tinker-review/playbooks/general.md +++ b/.skills/tinker-review/playbooks/general.md @@ -46,11 +46,22 @@ confidence pass so the `governs` edges they write go through it: - Test helpers without guard clauses (missing else/throw for invalid input) **Project memory** (only when beads were found; read decisions by `listBeads`): -- Code against chosen decisions — does the diff do something a decision's design +- Chosen decisions are claims to verify, not reasoning to accept. A design + states facts — a mechanism, an invariant, a measurement, what the choice rules + out. Check each against the code, and against the build or the functional test + where the claim is about behavior. Does the diff do something the decision rules out? Judge against the newest decision in a `supersededBy` chain. -- Code against rejected alternatives — each one is a "why not X?" already - answered; don't raise it again. But if the diff *does* X, check whether the - reason X lost still holds in this code. +- Rejected alternatives — each answers "why not X?" only if the reason X lost + is true of this code, so check that reason the same way. If the diff does X + anyway, does the reason still hold? +- Where the beads stop, look harder. Changed code no decision governs (no + `governs` edge after `linkBead`), and any part a bead says was built without + beads (a decision accepting that earlier commits have none), had its + reasoning recorded nowhere. Give that code the closest read, not a note. +- Trade-offs argued only in the PR thread — read `discussions.prComments` and the + JIRA comments for design debates that ended in a choice (concurrency, + cardinality, compatibility, ...). Each one with no decision bead under the + PR's root is reasoning that lives only in a comment thread. - `in_progress` tasks under the PR's root against what the PR says it delivers — claimed work that isn't here, or is here half-done. - Decisions that record a departure from the JIRA, proposal or dev@ thread — @@ -110,6 +121,16 @@ author's recorded reasoning is as much inside knowledge as the source code. - `discussions.beads` — the root is expected to be open: it closes at merge, so an open root is never a finding, and neither is finding no beads or a failed sync (the report says which). +- A chosen decision's claim that doesn't hold in the code or the build — high: + the change rests on it. A rejected alternative whose stated reason is false + of this code — an open question: the road may have been closed for the wrong + reason. +- A trade-off settled in the PR thread with no decision bead — an open question + per trade-off, naming it and citing the comments, so it can be recorded before + merge. Raise it only when the PR has beads; without any, the thread is the + only record and that is not a finding. +- Code where the beads stop is not a finding by itself; weigh what the closer + read turns up like any other Inspect candidate. - Diff does what a rejected alternative ruled out and the reason it lost still holds — high. If that reason no longer holds, an open question: was this a deliberate reconsideration? Diff contradicts a chosen decision with no newer
