What problem does it solve? Self-modifying AI skill systems need a hard boundary between proposing changes and landing them; without it, a loop could apply its own patches or trust forgeable verdicts. This Skill is the only privileged gate that mutates the skill corpus, enforcing human authorization plus a fresh cross-model jury verdict on every staged patch. ## Core Features & Use Cases - Jury-at-landing verification: Runs a fresh cross-model codex jury on each staged diff at landing time, ignoring any producer-written verdict that could be forged. - Privileged corpus mutation: Applies approved patches via Write/Edit with automatic backups to .aris/meta/backups/ and provenance stamping (author, reviewer, verdict ID, content hash). - Audit logging and tracing: Records every landing to .aris/meta/optimizations.jsonl and saves jury traces for forensic recovery. - Use Case: After running /meta-optimize and reading its report, you type /meta-apply 1,3 to land only the patches you named, each re-judged by a fresh codex jury before any file changes. ## Quick Start After reviewing the staged patches from a /meta-optimize run, ask the assistant to run /meta-apply with the patch numbers you want to land.
