Use this at the end of substantial repo or agent waves to convert evidence-backed lessons into proposed durable assets: skills, scripts, rules/checks, prompt templates, docs, or issues. Always use it when the user mentions flywheel, wave closeout, repo ecosystem learning, durable asset promotion, or learning-to-tools.
Recover and harden digitalmodel Blender automation work in isolated worktrees when uv/editable dependency paths break and local machines lack a Blender executable.
Class-level digitalmodel OrcaWave/OrcaFlex readiness, semantic-proof, fixture-proof, and closeout workflows.
Run digitalmodel tests from isolated worktrees without uv editable-dependency failures by using the main repo's existing virtualenv and PYTHONPATH.
Create evaluation scripts and integration tests for Python scientific libraries in the digitalmodel package. Follows the established pattern from fluids, ht, meshio, sectionproperties, and pygmt evaluations.
Build and extend fixture-backed OrcaFlex reporting proof paths in digitalmodel using stable metadata baselines, normalized HTML snapshots, and reusable reporting test helpers.
Architecture and scripts for syncing Hermes memory into git-tracked .claude/memory/ so all machines get context via git pull. Covers quality gate, drift detection, topic mirroring, and cron automation.
Operate and recover the Hermes-to-repo memory bridge: drift checks, quality gate, bridge commits, push verification, and stash recovery when pre-bridge scripts fail after generating outputs.