Skip to content

case G: a scoped review of the three leaves whose mutants survive - #47

Merged
buzzdan merged 1 commit into
mainfrom
claude/case-g-leaf-mutants
Oct 18, 2026
Merged

buzzdan merged 1 commit into
mainfrom
claude/case-g-leaf-mutants

Conversation

@buzzdan

@buzzdan buzzdan commented Oct 15, 2026

Copy link
Copy Markdown
Owner

What

A new scoped case, case-g-leaf-mutants-review, in go, py and ts-react. Its prompt reviews the three files that hold the R7.Q7 plants: the device id parser, the tenant type and the strings helpers.

Graders:

  • Three recall graders. Each one needs the file and the row the surviving mutant needs, on the same line:
    • the 64-character row
    • the tenant rows: the rune range in Go, the error message in py and ts-react
    • slugify's separator rows
  • q7-judged: fails a report that leaves Q7 per-leaf on this scoped review.
  • no-edit, no-write, no-multiedit, rule-cited (copied from case C) and evidence-protocol (per language).

Why

On a whole-repository review, R7 Q7 is now a per-leaf question that routes to a scoped review (buzzdan/ai-coding-rules#98). Case G measures that scoped review.

Checks

  • evalcase.Discover loads all three suites from the copied layout (cases/ + scaffold/), with case G in each.
  • Each recall pattern matches a sample finding row and does not match an unrelated row.
  • No baseline yet. Next: 2 runs per suite (~$4 each).

🤖 Generated with Claude Code

R7 Q7 is a per-leaf question: a whole-repository review leaves it unjudged
and routes it to a scoped review. Case G is that scoped review, in every
suite: the deviceid parser, the tenant type and the strings helpers, the
three files that hold the R7.Q7 plants. Each recall grader names the row the
surviving mutant needs, on the same line as the file, not the file name
alone: the 64-character row, the tenant rows (rune range in Go, the error
message in Python and ts-react), and slugify's separator rows. q7-judged
fails a report that leaves Q7 per-leaf on this scoped review.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@buzzdan
buzzdan merged commit 5feabd9 into main Oct 18, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant