Headless-friendly regenerate → score → report loop for OpenClinXR factory/evidence work. Use for model-vetting, orchestrate_character, garment/skin regen, or any bounded regen-then-measure cycle under OPENCLINXR_WORKER=1.
Install
npx skillscat add simnova/openclinxr/regen-score-report Install via the SkillsCat registry.
SKILL.md
Regen → score → report (headless)
Cheap, deterministic loop for workers. Parent sets scope + OPENCLINXR_WORKER=1; worker does not update SSOT registries or PROJECT_STATUS.
Preconditions
export OPENCLINXR_WORKER=1
export OPENCLINXR_JOB_TMP="${TMPDIR:-/tmp}/openclinxr-job-$$"
mkdir -p "$OPENCLINXR_JOB_TMP"
# worktree cwd with pathScope writeRoots onlyLoop
- Regen (bounded CLI only; unique outs under job tmp or worktree artifact dir)
- Character/pipeline: repo
orchestrate/ package scripts with explicit--out/ job-tmp - Never fixed
/tmp/openclinxr_*basenames (per-job-tempskill)
- Character/pipeline: repo
- Score / verify (code first, not LLM)
- Focused package tests or existing report packers
- Pixel/schema gates where available
- Report
- Write handoff / artifact-map under
.openclinxr/slices/<id>/handoffs/<role>.jsonor assigned write root - BLUF + file:line evidence paths; no PROJECT_STATUS append (parent integrates)
- Write handoff / artifact-map under
Headless invocation
OPENCLINXR_WORKER=1 grok -p "Follow .grok/skills/regen-score-report: regen X → score → write handoff only" \
--model deepseek-v4-pro --yolo --cwd <worktree> --max-turns 30 \
--output-format jsonAnti-patterns
- Running
docs:hygiene/ registry rewrites in the loop - Sharing one Model Vetting output dir across parallel jobs without unique subdirs
- Calling frontier model for pure regen when deepseek-pro + scripts suffice