feat(skills): evidence-bar false verdict; done means the tests discriminate - #2816
Draft
pbean wants to merge 1 commit into
Draft
feat(skills): evidence-bar false verdict; done means the tests discriminate#2816pbean wants to merge 1 commit into
pbean wants to merge 1 commit into
Conversation
…minate Two triage-economics changes. 1. A finding that names no bad outcome, or no input or state that reaches one, is `false` with "no demonstrated outcome" as the refutation. Triage does not build the claim on the reviewer's behalf. Applied at all four triage sites. 2. Step-03 in bmad-build and bmad-build-auto adds two things to done: a test this change added must fail with the change reverted, and a helper this change introduced must be adopted at every site the change touched. Findings the verification-gap layer would otherwise raise at review are closed before review starts.
pbean
force-pushed
the
feat/triage-evidence-bar-and-ledger
branch
from
September 2, 2026 01:20
1c13775 to
3f5e4ad
Compare
pbean
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
Two places in the review pipeline spend effort on work that produces nothing.
Triage verifies every finding a layer reports, including the ones that make no claim — a "consider extracting this", a "this could be simplified". There is no bad outcome to check, so triage either invents one on the reviewer's behalf or rejects it with a reason it has to compose. Either way it pays for a finding that was never a defect.
Verification gaps reach review that step-03 could have closed. A test the change added that would still pass with the change reverted, or a helper the change introduced that a touched site does not use, is not done — but step-03's definition of done did not say so, so the gap layer raised it, triage verified it, and a patch pass fixed what the implementer could have finished in place.
What changed
Instruction text only, six files. The two step-04 review files stay in lockstep.
A finding with no demonstrated outcome is
false. At all four triage sites —bmad-buildstep-04 and one-shot,bmad-build-autostep-04,bmad-code-reviewstep-03 — a finding that names no bad outcome, or no input or state that reaches one, getsfalsewith "no demonstrated outcome" as the refutation. Triage does not build the claim for the reviewer. The row still exists; it just costs nothing to write.Done means the tests discriminate. Step-03 in
bmad-buildandbmad-build-autoadds two things to done: a test this change added or changed must fail with the change's production code reverted (or under the wrong behavior it exists to catch, when the change is the test itself), and a helper or chokepoint this change introduced must be adopted at every site this change touched. Judge from what the test asserts, or revert and run when that is cheap. When finishing either changes code, the diff file is rewritten and re-read.Verification
npm testgreen: renderer 24/24, skill validator, lint, markdownlint, prettier.