Scripts that are part of protocol but were never invoked in real engagement traffic. They work — but their value-add wasn't validated under real load. Kept here as opt-in, not auto-invoked.
| Script | Trigger to use it |
|---|---|
engagement-doctor.py |
Manual diagnose when engagement state looks corrupted; --check stalled if heartbeats absent for >threshold |
engagement-migrate.py |
One-shot migration of legacy engagement (no protocol_version, missing iteration counter, etc.); --include-archived for archived ones |
token-budget.py |
Tier-budget guard (S=100k / M=500k / L=1.5M) |
Promote back to scripts/ when one of these gets used in ≥2 real engagements with measurable value.
The following optional scripts were removed in an earlier cleanup:
cross-val-template.py— never invoked in live skills; cross-val table is now written inline by the lead in handoff §4.director-sweep.py— orphaned after the manager/director split; managers don't re-run validator sweeps (same brain = no new info).validator-retry.py— logic absorbed intovalidator_lg.py's retry edge.secondary-init.py— was never wired into any cross-domain handoff flow.metrics-summary.py— no live caller; retrospective metrics deferred to a future LangSmith dashboard.
size-detect.py was promoted from optional/ to scripts/ root — it's actively invoked by handoff-precheck.py and three protocol skills (agency-intake, engagement-protocol, acceptance-protocol), so the optional/ categorisation was wrong.