Repository navigation
Description
Parent
Refs #15703
Source: #16029, part 2 of #15729. The list recorded in #16602 gives /instruction-placement:audit ("routing rubric's own ladder") and /instruction-placement:realign ("applies the audit's routing rubric") the route "call /corrections:codify, presence-gated, with the strongest-first fallback".
Depends on
- none open. The instruction-placement cut (Cut the model-read ecosystem specifics from instruction-placement #16765) already landed its routing-rubric edit through the merged fix(instruction-placement): cut model-read ecosystem specifics #16975, so nothing pending touches these lines.
Outcome
Today, when the routing rubric finds that a candidate is mechanically checkable (rung 2), it chooses the mechanism from its own list ("a compiler, linter, formatter, analyzer, or hook"). After this change the rubric hands that choice to /corrections:codify when it is available, and falls back to the shared strongest-first sentences when it is not. Both skills keep reporting the route without building it, and keep the prose until the mechanism exists. One rubric edit covers both skills, because audit and realign read the same file.
Demo: run /instruction-placement:audit on a repository whose CLAUDE.md holds a formatting rule a check could decide, with a corrections stand-in plugin loaded. The finding names /corrections:codify as the next step for that rule, does not pick a mechanism itself, and does not mark the prose for deletion.
Where the text changes
Line numbers as read on main at a483a2c1; they may move.
| File and lines | Present wording | Change |
|---|---|---|
plugins/instruction-placement/context/routing-rubric.md:83-86 (rung 2 of "The decision ladder") |
"A compiler, linter, formatter, analyzer, or hook could enforce it deterministically → route to that mechanism. Formatting rules, import ordering, naming patterns a regex can express. Never send an agent to do a linter's job. This plugin reports the routing; it does not author linters, and it does not delete the prose until the mechanism exists." | Replace the mechanism list with the gated codify call plus the shared fallback (below); keep "reports the routing" and "does not delete the prose until the mechanism exists" |
plugins/instruction-placement/context/routing-rubric.md:225 ("What this rubric does not decide") |
"Authoring the linter, hook, or skill that rung 2 and rung 3 route to." | Say that choosing and proving rung 2's mechanism is /corrections:codify's, when available; authoring stays out of scope |
plugins/instruction-placement/skills/realign/context/apply-recipes.md:185-187 ("Report-only outcomes") |
"Rung 2, mechanical enforcement. The remedy is a linter, formatter, analyzer, or hook. Report the routing and leave the prose in place." | The blocked finding names the codify call (or the fallback) instead of a mechanism; the prose still stays in place |
plugins/instruction-placement/skills/audit/context/routing-out.md:43-46 and plugins/instruction-placement/skills/realign/SKILL.md:258 |
"content a linter should enforce"; "A linter or skill routing has no destination this skill" | Reword only where they name a mechanism for rung 2; no behavior change beyond the rubric's |
The call form: /corrections:codify rules <the instruction file> for the candidate's own file, or /corrections:codify "<the rule>", whichever codify's argument hint fits; the audit names the call in the finding and never invokes it unasked (the audit is read-only).
Decisions
-
Design Q21 (Resolve ecosystem specifics from the consumer repository and align the enforcement ladder with the talk's pyramid #15703): a skill that carries no ladder copy reads the ladder by calling codify; only a skill that applies its own selection rule holds a synced copy and a Selection rules row.
-
docs(enforcement-ladder): add the architecture and retro Selection rules rows in ladder 3.2.0 #16602 recorded both skills under "call
/corrections:codify, presence-gated, with the strongest-first fallback (12; change, filed as follow-on items by part 2, List enforcement-placing skills and give each one route to the ladder - part 2 of 2 #16029)". -
The fallback wording is shared by all nine follow-ons of List enforcement-placing skills and give each one route to the ladder - part 2 of 2 #16029, so the skills do not drift apart. After the gate sentence ("when
/corrections:codifyis among the available skills, hand it the choice of mechanism; it picks the strongest rung of the enforcement ladder that can decide the rule and proves the check"), the fallback is exactly:Otherwise, propose the strongest mechanism that can decide the rule: a change that makes the mistake impossible first, then a check that blocks it, then a test, and an instruction only when nothing a machine runs can decide it. Say that
/corrections:codifywas not available, so the proposal is unproven.
Key interfaces
- "A skill that carries no ladder copy calls
/corrections:codify, presence-gated, with a two-sentence strongest-first fallback. Only a skill that applies its own selection rule holds a synced copy and gets a row in the ladder's Selection rules table." (List enforcement-placing skills and give each one route to the ladder #15729) - The seam-phrasing convention: the gate is an availability condition ("when it is among the available skills"), the fallback is stated beside it, and the reference names the plugin, never the marketplace.
- The fallback names no tool and no ladder file path: ADR 0101 forbids a shipped tool choice, and these skills carry no ladder copy.
Done when
- Rung 2 of the routing rubric hands the mechanism choice to
/corrections:codifybehind the availability gate, followed by the shared fallback quoted under Decisions, and no longer lists mechanisms of its own. - Realign's "Report-only outcomes" names the same route for a rung 2 finding, still marks it
blocked, and still leaves the prose in place. - Neither skill gains a ladder copy or a Selection rules row, and no rung list is restated.
- The pull request body quotes, for each skill, the sentence that picked a mechanism from a private list before the change. If the recheck finds none for a skill, that skill is left unchanged and the finding is reported on List enforcement-placing skills and give each one route to the ladder - part 2 of 2 #16029.
- One new eval case per changed skill (audit, realign) loads a corrections stand-in plugin, in the shape of session-flow's
re-mechanical-finding-check-proposalcase, and grades that the finding hands the checkable rule to/corrections:codify, picks no mechanism itself, and keeps the prose. - The pull request body names, for each changed skill, the existing case whose run reads the changed text, or says none does. The lane runs no eval.
- A changelog fragment under
.changes/instruction-placement/describes the change. - AC10: no new enforcement-placing skill is added here.
- Open the pull request with
Closesthis issue in its body, then stop: the lane adds no label.
After the pull request opens (host session, not the lane)
The pull request stays held with do-not-merge for the eval; a host session runs only the eval and posts its record; the merge lane releases the hold when the record matches the head.
- Hold line: the hub's sweep (or a host session) adds
do-not-mergeand posts, from a listed human's account, a comment with the line<!-- pr-hold reason=eval -->on a line of its own. - Run budget. List enforcement-placing skills and give each one route to the ladder #15729 asks for "that skill's cases on the
opusandsonnetaliases, three runs per side, and no more than 240 runs per plugin". The owner decision of 2026-11-16 in the owner-decisions record (hop 4 Q15, which names List enforcement-placing skills and give each one route to the ladder #15729, amended by hop 7 Q3) replaces the run counts: per changed skill, one targeted before-and-after case onsonnetat 3 runs a side;opusruns only to confirm a sonnet loss; a new case gets a 1 to 3 run smoke first. Select cases with--case, use--runs 3and a--max-cost-usdceiling, and start no batch after a usage-limit error. Ceiling for this plugin: 2 skills x (3 smoke + 6 sonnet + 6 opus confirm) = 30 runs, under the parent's 240. - Control: the removed mechanism list is a partial cut, so the pull request body carries a detection-control result. No paired no-loss compare (#15703 comment 6906800940).
- On a LOSS reading: a trace read of each dropped case and an
opusconfirm; only a loss traced to the changed text counts. - Record: the verdict and the control result go into the pull request body. When the run is VALID and shows no loss, the host session posts, from a listed human's account, a comment with the line
<!-- eval-pair verdict=no-loss validity=VALID head=<40-hex head SHA> -->naming the current head, keeping "do not merge" anddo-not-mergeout of that comment (docs/contracts/loop-lane/README.md, "Cross-lane PR hold"). A loss or a run that is not VALID leaves the hold in place; say so on the pull request. - Release: pr-merge's scheduled step "Release eval holds whose record matches the head" removes
do-not-mergeonce that record names the current head and no other hold stands. A push after the record needs a new record.
Out of scope
- Building any linter, hook or check the route names.
- Rung 3 (skill) routing, the hard-deny gate, glob derivation and the promote lane.
- Every other plugin's follow-on.
Related
- Resolve ecosystem specifics from the consumer repository and align the enforcement ladder with the talk's pyramid #15703 (container)
- List enforcement-placing skills and give each one route to the ladder #15729
- List enforcement-placing skills and give each one route to the ladder - part 2 of 2 #16029
- docs(enforcement-ladder): add the architecture and retro Selection rules rows in ladder 3.2.0 #16602
Metadata
Metadata
Assignees
Labels
Type
Fields
Work status
Priority
Activity
github-actions commentedon Nov 18, 2026
Size: Medium
Complexity: Medium
Uncertainty: Known
Flags: none
Blocked by: none
Blast radius: instruction-placement's routing rubric, the realign apply recipes, audit and realign skill text, two new eval cases and a changelog fragment; read by anyone running /instruction-placement:audit or :realign.
Reversible: yes; revert the text edits, eval cases and fragment.
Open questions: none
Understood: rung 2 of the routing rubric should hand the mechanism choice to
/corrections:codifywhen it is available, with the shared strongest-first fallback otherwise. Realign's report-only outcome should name the same route, and the prose should stay in place. The issue gives exact wording, the files to change and the done-when list, so nothing is missing. The sibling follow-ons (#17179, #17175, #17174, #17172, #17162, #17160, #17159, #17158) cover other plugins, so none is a duplicate. I found no merged pull request that already does this.Automated triage. A maintainer reviews it.
melodic-automation-lanes commentedon Nov 18, 2026
Not admitted to the issue lane: its done-when asks for what an unattended lane run cannot perform: item 4 (tracker-output: "The pull request body quotes, for each skill, the sentence that picked a mech..."), item 5 (eval: "One new eval case per changed skill (audit, realign) loads a corrections stan..."); recommended split: keep the other items here as the admissible issue and move these to a child issue for the host queue (
host-run). Admission takes the issue once each of those is met, when intake triages it again (.github/actions/resolve-config/admission-rules.json).melodic-automation-lanes commentedon Nov 18, 2026
Refinement. A read-only run explored this repository, searched its issues and pull requests, and read the allowed docs to draft what follows. On an issue a listed human filed, it counts as the issue's spec once posted, while unedited; it is still model text, so check it.
Intended outcome: Rung 2 of the instruction-placement routing rubric hands the mechanism choice to /corrections:codify behind an availability gate with the shared fallback, and realign's report-only outcome names that route while staying blocked and keeping the prose.
Split. Done-when items 4, 5 of the issue body need a person: they are checks a person runs, listed below, and the lane does not do them. This issue keeps the rest of its done-when.
Not done by the lane: checks a person runs
External state: none
Related work
Sources read
Automated refinement by the lanes App. Intake triages the issue again with it.
github-actions commentedon Nov 18, 2026
The triage sweep dispatched intake triage for this issue (dispatch 1; 2 runs that reach intake, or 6 dispatches, hand it to a maintainer).
melodic-automation-lanes commentedon Nov 18, 2026
{"v":1,"actor_id":337595936,"title_sha256":"78db5de65a504899d35e45a3c645132a5264f9c80c266815b73602448164adc9","body_sha256":"e2a3e63b439da0ac1d445830e922bd42eb2bead4348d6c4efaa4186c3c9aca29"}
melodic-automation-lanes commentedon Nov 18, 2026
{"v":1,"actor_id":337595936,"title_sha256":"78db5de65a504899d35e45a3c645132a5264f9c80c266815b73602448164adc9","body_sha256":"e2a3e63b439da0ac1d445830e922bd42eb2bead4348d6c4efaa4186c3c9aca29"}
github-actions commentedon Nov 18, 2026