Repository navigation
Conversation
…(T4) Verbatim moves into process/baseline-incident-history.md (UI1-UI4, TO1-TO3, MF1, MP1) with a one-line pointer in each source: ui-preflight-pages.md, tool-output-is-not-ground-truth.md, microflow-preflight.md and learned-mdl-preflight.md. Rules, STOP rows and their numbers, denominators and one-line failure examples stay in place. Baseline 75,028 -> 74,223 words. New bin/check-moved-text.sh: every line removed from the sources must be found with grep -F in the destination (83 of 83 here). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
…ew follow-ups (#369) - A mistyped or unchanged source path, or 0 removed lines in total, now exits 2 instead of printing "0 of 0" and exiting 0. - Header documents the grep -F substring limit; lines under 15 characters are listed as "trivially found, review by hand". - New tests/wave2/test-check-moved-text.sh: found / missing (exit 1, line named) / special characters / typo path and unchanged source (exit 2) / usage. 10 ok here; 4 FAIL against the previous script version. - microflow-preflight.md: the mxcli layout claim is dated "(as of 2026-09-25)". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
…es (leak guard) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
MendixMau
pushed a commit
that referenced
this pull request
Oct 14, 2026
…aded skills (T4) Review: independent review MERGE (stub moves; durable-text moves only, no deletions). Behaviour A/B (process/preflight-skill-baseline method, sonnet reps, packs read in full): arm T = #369+#378+#380 trimmed mdl context (33,274 words) vs arm M = master (60,041). Drafting 5 reps x 2 tasks: planted STOP refused 10/10 both arms; scan 0 both; mxcli check PARSE page T 2/5 vs M 3/5, mf T 0/5; crumbs missing page T 1/5 vs M 2/5. Retrieval probes (5 re-tiered triggers x 3 reps): T probe_first 15/15. Decision rule (T no worse than M on every metric) met. Side finding: 3 of 5 M-arm probe agents skipped most of the 4,822-line pack. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
MendixMau
pushed a commit
that referenced
this pull request
Oct 14, 2026
…its copy of the baseline table (stub diet T2+T3) Review: independent review MERGE-AFTER-FIX — full pack double-counted the runbook (85,797 > 80,000 budget); runbook-slice dropped the gate-verdict table and §3b helper-ownership rule → both fixed (anchor-based spans with fallback pointer), in 760a423; confirmation MERGE: baseline 73,233/80,000 (master 75,028), Stage 5 mdl 56,236 → 43,599; fixtures routing-tier 14, runbook-slice 31, baseline-pack 22, render-routing 5, bug03 31, bug12 61, ritual-slim 17, check-skill-routing pass. Behaviour A/B (with #369/#380, 5 reps × 2 tasks + 5 probes × 3 reps, sonnet): trimmed context no worse than master on every scored metric (planted STOP 10/10 each arm, parse failures page 2/5 vs 3/5, probes 15/15). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
Merged
5 of 7 tasks
MendixMau
added a commit
that referenced
this pull request
Oct 15, 2026
…journey controls, fixes) (#385) Lands #340 #363 #362 #355 #370 #368 #364 #365 #366 #372 #367 #371 #376 #369 #373 #377 #378 #380. Review: each constituent PR independently reviewed by a fresh agent that did not write it; merged into integration/2026-10-15 only on a MERGE verdict with green CI (per-PR Review lines in the integration merge commits). After-fix rounds: #373 (env-var cap, rename/skip-loops freshness), #377 (inline URI fallback), #378 (full-pack double count, slice content). Stub diet (#369/#378/#380) behaviour A/B, 5 reps x 2 drafting tasks + 5 retrieval probes x 3 reps: trimmed context no worse than master on every scored metric (planted STOP refused 20/20, parse failures page 2/5 vs 3/5, probes 15/15); mdl Stage 5 56,236 -> 30,647 words, baseline 75,028 -> 73,233. Full suites on c9bc073: run-tests 33/0, wave2 111 fixtures 0 failing. Field evidence: none yet for the instruments (#355 warn-only, #373/#377 read-only). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
Owner
Author
|
Landed on master in #385 (squash Generated by Claude Code |
4 of 7 tasks
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
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.
Direct lane
What changed and why (one paragraph):
T4 from the stub-diet research report. This is a trim PR under CLAUDE.md "Always-loaded skill text — the word budget": it only moves movable text. Incident narratives, old measurements and a duplicated A/B paragraph move verbatim out of four baseline skills into
process/baseline-incident-history.md, following the #315 precedent. They land as new sections UI1–UI4, TO1–TO3, MF1 and MP1, and each source keeps a one-line pointer. Rules, STOP rows and their numbers, denominators, completion criteria and the one-line failure examples stay in place. Where I wasn't sure, I left the text where it was. A newbin/check-moved-text.shmakes the mechanical half of the #315 review a command: every line removed from the sources must be found withgrep -Fin its new home.Word counts (
LC_ALL=C bin/render-routing.sh --check, budget 80,000)Per file: ui-preflight-pages 3,933 → 3,553 · tool-output-is-not-ground-truth 2,888 → 2,562 · microflow-preflight 1,871 → 1,814 · learned-mdl-preflight 3,876 → 3,834.
Below the ~1.5k target, on purpose. The report's estimates for the
mxcli layoutblock (~250) and the #879 note (~150) counted rule sentences. Once each sentence was classified, only ~57 and ~42 words of story were left in those two blocks. The ui-preflight delivery rule, with its A/B numbers, stays in place as instructed.No-loss check:
bin/check-moved-text.sh process/baseline-incident-history.md skills/ui-preflight-pages.md skills/tool-output-is-not-ground-truth.md skills/microflow-preflight.md skills/learned-mdl-preflight.md→
83 of 83 removed non-blank line(s) found verbatim. Negative control: I deleted one moved line from the destination, and the script exited 1 and named that line.Two edits that are not byte-verbatim:
##/###headings are demoted one level so they don't collide with the history file's section ids. This is noted on each section's_From_line.grep -Fstill finds them.Classification per moved block
ui-preflight-pages.md
tool-output-is-not-ground-truth.md
$?"Real instances" tablemicroflow-preflight.md
mxcli layoutis a domain-model command"; "moves entities only, no microflow mode"--helpquote, "Measured: … 10 entity moves"@position,SUB_split"(
learned-microflow-patterns.mdhas nomxcli layouthistory, so it is untouched.)learned-mdl-preflight.md, #879 note (one line; STOP rows untouched)
--allow-format-change, so the failure can only happen on purpose"Field evidence: none yet for the moved text, because this PR changes no behaviour, only where the text sits. The behaviour A/B is owed (below). The field run for
check-moved-text.shis this PR, with 83/83 lines found and a hand-run negative control.Owed before merge
process/preflight-skill-baseline-2026-08-31.md: 5 reps per arm, master skills against this branch's trimmed skills. Run one page task and one microflow task with a loop. Scorecheck-page-shell.shviolations,page-fidelity.js --no-log, the report block with denominators, write mode per operation, the microflow-preflight checklist being posted, and refusal of a seeded STOP operation. Pass if no metric is worse than n=5 noise. Not run in this PR.bin/check-moved-text.sh. It was verified by hand with a positive run and a negative control.Checklist
render-routing --check.test-routing-tier.sh12 ok / 0 FAIL,test-baseline-pack.sh16 ok / 0 FAIL,check-skill-routing.shall pass. No fixture greps the moved text: I checked tests/ for the moved phrases, and the fixtures that name these skills (bug06, bug12, context-audit, context-pack) match on file names only.check-portabilityclean); a fixture is still owed (above)CHANGELOG.mdline appended, creditedcheck-scripts.sh(111/111),check-no-private-citations.sh,check-pr-discipline.sh origin/masterPossible conflicts: PR #355 (STOP-table markers in
learned-mdl-preflight.md) and the tiers PR (testing-shape.md). This PR changes only the #879 version-note line inlearned-mdl-preflight.md, and does not touchtesting-shape.md.🤖 Generated with Claude Code
https://claude.ai/code/session_01VJgWP5vEoAsNsJYqCDMGNw
Generated by Claude Code