workflow_lint size-ratchet: 09-step-5.md + 18-step-10d.md regrew past SKILL_DOC_SIZE_GRANDFATHER caps — fleet-wide no-flags rc=1
Highlight text on any card (body, plan, or an event) to anchor a comment, or leave a whole-task comment here. Mention @claude to summon a reply.
No comments yet.
Goal
Restore a green no-flags workflow_lint.py run fleet-wide: two main-tree skill step docs regrew past their SKILL_DOC_SIZE_GRANDFATHER ratchet caps without the #1753 landing-bytes cap raise, so every no-flags lint run (inline payload lint gates, Step 9c-adjacent runs, pre-push legs) now exits rc=1 with 2 spurious errors that sessions must hand-attribute as pre-existing.
Symptom (measured 2026-08-20, worktree fresh-synced to origin/main)
.claude/skills/issue/steps/09-step-5.md: 103,751 B > ratchet cap 100,300 B.claude/skills/issue/steps/18-step-10d.md: 287,529 B > ratchet cap 282,700 B
Some landing on these files skipped the ratchet protocol (the lint's own remedy text prescribes a measured cap raise at landing time).
Fix
Per the lint's remedy text, either raise the two SKILL_DOC_SIZE_GRANDFATHER caps in scripts/workflow_lint.py to the measured size + the ≤2.8 KB corridor, or trim the docs back under their caps. Verify with a no-flags workflow_lint.py run exiting 0 (given no other new red), and run the lint-family pin tests (tests/test_workflow_lint*.py).
Provenance
Surfaced by the round-6 experiment-implementer on task #2378 (payload-attribution: the 2 errors name files byte-identical to origin/main, no round-committed file). Dedup checked 2026-08-20: no existing proposed task covers these two caps (#2252/#2327/#2374/#2409/#2414/#2420/#2421/#2423/#2426 are adjacent but distinct).