Treat repeated failures as class evidence and reopen the kernel after compaction (4.8.4)

[mzored]

Summary

• The always-loaded kernel now treats a second appearance of the same kind of failure (an owner reporting a symptom again, a slow gate failing on one more stale expectation) as evidence about a class: find the other members with the cheapest check that covers the whole class and fix them together before returning to the loop that found them.
• The bug workflow's sibling inspection also covers the other places the owner would see the same symptom, since a repeated symptom can have a second cause.
• The skill description carries one conditional sentence: a session this skill was governing reopens the skill after the host compacts its context. Not a hook; executes nothing. It answers the reopen clause of the 3.0.0 reminder decision, which fired in a private installed Codex session on 4.8.2 where no package file text survived any of fifteen compactions.
• Version 4.8.4; CHANGELOG, docs/evidence.md, and docs/decisions.md record the observations, prior-art reading, refused alternatives, and what stays UNVERIFIED.

Validation

• scripts/check.py: 389 tests passed; git diff --check clean; both host schema validators passed.
• Isolated read-only Codex review: one qualifying finding (a contradiction between "none of the package's text survives" and "the description is package-owned text"), fixed; no findings refused.
• Model behavior under both wording changes stays UNVERIFIED by design; no synthetic receipts were run.

🤖 Generated with Claude Code