Skip to content

Cut the model-read ecosystem specifics from code-tidying: batch-simplify, tidy-diff and audit-dead-code #17755

Description

@kyle-sexton

Part 1 of 5 of #16756. Refs #15703. Blocks #15732 through #16756.

Problem

Three code-tidying skills read a per-language adjudication table, a manifest detection list and examples written in one stack's syntax.

Rows (4: all R11)

Row Rule Kind Cited at (as of 5b66a7f) Fact
F0092 R11 K4 plugins/code-tidying/skills/audit-dead-code/context/adjudication.md:7-32 evidence-pattern table (MediatR, Flask, click, pytest, reflection idioms)
F0128 R11 K2 plugins/code-tidying/skills/batch-simplify/context/reference.md:11 group by nearest *.csproj, package.json, pyproject.toml, Cargo.toml
F0129 R11 K7 plugins/code-tidying/skills/batch-simplify/SKILL.md:326 "a .sh hook that invokes ruff"
F0130 R11 K7 plugins/code-tidying/skills/tidy-diff/reference/generated-code-residue.md:42-89 detection cues in TS/JS syntax (as any, @ts-ignore, console.log)

Files

  • Cut: plugins/code-tidying/skills/audit-dead-code/context/adjudication.md, plugins/code-tidying/skills/batch-simplify/context/reference.md, plugins/code-tidying/skills/batch-simplify/SKILL.md, plugins/code-tidying/skills/tidy-diff/reference/generated-code-residue.md.
  • Outside references at d46efa9: .performance/ratchets.json:172 (adjudication.md skill-stack-fences ceiling 2). audit-dead-code ships scripts/dead-code-scan.sh and lib/dead-code-shapes.sh; the script-reader search must show whether either reads adjudication.md.

Rules this part follows (from #16756, rules revision 5)

  • R11 (shipped knowledge): model-read K2, K3, K4, K5 or K7 text. Generic text replaces it only at a routing site where the skill still takes a language-specific step; anywhere else it is deleted with no replacement. An item kept only because a removed item consumes it goes in the same pull request.
  • R4 (consumer default): the value picks a tool or library for the consumer. It goes: no bundled default survives (ADR 0101 decision 2, "Ship no bundled default"); the choice is resolved from the consumer through docs/contracts/ecosystem-resolution/.
  • R8 (measured fact): cut it like the other rows and mark it in the pull request body as pending its targeted case; the host session's eval (below) decides whether it comes back, annotated portability-ok: measured <case-id>.
  • AC1's named exceptions are the kept classes of ADR 0101 (docs/adr/0101-resolve-ecosystem-specifics-from-the-consumer-repository.md, Decision and the 2026-11-17 amendment). Bundled defaults that pick a tool, command or library are retired wherever they appear (Interview Q32, D4 of the Brief).
  • Line numbers below were read at 5b66a7fa and may have moved; find each row by its cited text.

After the pull request opens (host session, not the lane)

The lane opens the pull request with Closes this issue and stops; it adds no label. The pull request stays held for the eval:

  • Hold line: a sweep or host session adds do-not-merge and posts, from a listed human's account, a comment with <!-- pr-hold reason=eval --> on a line of its own.
  • Targeted case per changed skill: one before-and-after case on sonnet at 3 runs a side, --case, --runs 3, a --max-cost-usd ceiling; opus only to confirm a sonnet loss; a new case gets a 1 to 3 run smoke first; start no batch after a usage-limit error (owner decision of 2026-11-16).
  • Controls: the owner's Q53 decision of 2026-11-17 (owner-decisions.md L202) reads: "each cut pull request is gated on a no-loss compare alone, with no wrong-text side". It supersedes Cut the model-read ecosystem specifics from code-tidying #16756's Controls line and its AC7 sentence ("each partial cut carries a detection-control result"). A whole-page deletion still states its predicted outcome in the pull request body before the first batch, and the record says the case cannot detect wrong text.
  • On a LOSS reading: a trace read of each dropped case and an opus confirm; only a loss traced to the changed text counts.
  • R8 rows: the row's targeted case with and without it; on a case drop or a mean fall of 0.15 or more without it, restore the row annotated portability-ok: measured <case-id>; a deterministic grader states before the run in how many of n runs it must fail.
  • Record: on a VALID no-loss run, post from a listed human's account <!-- eval-pair verdict=no-loss validity=VALID head=<40-hex head SHA> --> naming the current head, with no "do not merge" text (docs/contracts/loop-lane/README.md, "Cross-lane PR hold"). pr-merge's "Release eval holds whose record matches the head" step (.github/scripts/pr-merge/release-eval-hold.mjs) then removes do-not-merge; a later push needs a new record.

Out of scope

Rows outside this part's table, the parent's kept rows, every other plugin, toolchain's bundled defaults (#15733), and widening the portability gate (#15732).

Done when

  • Every row in this part's table is cut or replaced at a routing site; a row whose cited text is already gone on main is listed as GONE.
  • The pull request body states this part's row count (4) and lists every row id with its rule and outcome.
  • Each R11 row's script-reader search is quoted in the pull request body; a row whose file a script reads is not cut, and is listed in the pull request body under a Script readers heading.
  • The pull request body names, for each changed skill (audit-dead-code, batch-simplify, tidy-diff), the existing eval case that reads the cut text, or says none does.
  • The pull request body says, for each changed file, whether the cut deletes the whole page or part of it.
  • The skill-stack-fences ceiling for plugins/code-tidying/skills/audit-dead-code/context/adjudication.md in .performance/ratchets.json is lowered if the cut removes its fences.
  • AC10: no new enforcement-placing skill is added.

Activity

  1. added
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 21, 2026
  2. github-actions commented on Nov 21, 2026

    @github-actions
    Contributor

    Size: Medium
    Complexity: Medium
    Uncertainty: Known
    Flags: none
    Blocked by: none
    Blast radius: four model-read pages in the code-tidying plugin (audit-dead-code, batch-simplify, tidy-diff), plus the skill-stack-fences ceiling in .performance/ratchets.json. Anyone running those three skills reads them.
    Reversible: yes; revert the pull request, or restore a single row annotated portability-ok: measured <case-id>.
    Open questions: none
    I read this as part 1 of 5 of #16756 (refs #15703). It asks to cut or replace four R11 rows (F0092, F0128, F0129, F0130) at their routing sites. It also asks that the pull request body record the row count, script-reader searches, eval cases and whole-or-partial deletions per file, and that the ratchet ceiling be lowered if the fences go. The rules and done-when are exact, so I found no open design question. The post-PR eval hold is host-session work and is not part of this triage. I found no duplicate among the open issues, and #16756 is the parent tracker, not a duplicate.

    Automated triage. A maintainer reviews it.

  3. removed
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 21, 2026
  4. melodic-automation-lanes commented on Nov 21, 2026

    @melodic-automation-lanes
    Contributor

    {"v":1,"actor_id":337595936,"title_sha256":"245ab9a321480bb5b92561982d294a13c25cfb3fce209683caa442f987cc9339","body_sha256":"77dc0143678ef58954912f9ce05fa360c8bf184cfe1f0d577129786041232265"}

  5. github-actions commented on Nov 21, 2026

    @github-actions
    Contributor
    No description provided.
  6. github-actions commented on Nov 21, 2026

    @github-actions
    Contributor
    No description provided.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Fields

    Work status

    needs-triage

    Priority

    medium

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions