Methodology Gaps — Rolling Queue

Last reviewed: 2026-06-19
Frequency: weekly review (Sundays, synced with research synthesis)

This queue tracks claims in the clearinghouse directory that LAR cannot yet verify, either because no test exists or because the current environment cannot run one. Stale items (>30 days with no progress) are escalated.

Untested claims (no LAR test defined)

entry_id	field	first_seen	notes
openclaw	governance_hooks	2026-06-19	Needs test: kill parent process mid-inference, check session recovery
openclaw	memory_model	2026-06-19	Needs test: 1M token context stability under load
vllm	observability	2026-06-19	Needs test: scrape /metrics endpoint, validate OpenTelemetry export
ollama	governance_hooks	2026-06-19	Needs test: container restart behavior, model state persistence

Environment-limited (cannot test here)

entry_id	field	first_seen	environment_blocker	target_runner	owner	status	notes
microsoft-scout	runtime_type	2026-06-19	linux x86_64	Windows runner	unassigned	open	Needs Windows runner; openclaw Windows path already covered in Terminal v2026.6.8 post

Resolved

No resolved gaps yet. First resolution will be linked here.

---

How to close a gap

1. Write a LAR test that exercises the claim.
2. Run it with lar --from-directory <entry_id> --rerun.
3. Link the LAR run id in the resolved section.
4. Move the row from untested/environment-limited to Resolved.

Queue maintenance rules

• Every entry_id must resolve to an existing Markdown file in content/agents/.
• Phantom slugs are removed on discovery.
• Rows that reference a feature of an existing entry are rewritten to that entry's slug, or removed if the feature is already tracked elsewhere.