42 lines (35 loc) · 3.06 KB

Quality Score

Domain grades for Loong. Updated periodically to track gaps, prioritize cleanup, and measure harness maturity.

Domain Grades

Domain	Grade	Last Reviewed	Gaps
Contracts (L0)	A	2026-03-13	`#[non_exhaustive]` applied; membrane field not yet enforced at runtime
Kernel Security (L1)	B+	2026-04-03	Core tool paths now route through policy extensions; remaining gaps are non-uniform L1 coverage for connector/ACP/runtime-only analytics and explicit `Direct` compatibility lanes
Execution Planes (L2)	B	2026-03-13	Core/extension pattern solid; no WASM fuel metering yet
Orchestration (L3)	B	2026-03-13	HarnessBroker routes correctly; context-engine selection is pluggable, but richer engine implementations and broader runtime coverage are still limited
Observability (L4)	C+	2026-04-03	Durable JSONL/fanout audit exists; remaining gaps are tamper-evident verification, query baseline, and richer export lanes
Vertical Packs (L5)	B	2026-03-13	Pack validation works; namespace struct exists but not enforced
Protocol (L5.5)	B+	2026-03-13	Transport contracts and typed routing operational
Integration (L6)	B	2026-03-13	Plugin scanning works; hotplug lifecycle incomplete
Plugin IR (L7)	B-	2026-03-13	Bridge inference works; multi-language support limited
Self-Awareness (L8)	B-	2026-03-13	Snapshots generated but not continuous; no drift detection agent
Bootstrap (L9)	B	2026-03-13	Activation plans work; no policy-bounded bootstrap validation
Context/Memory	B	2026-04-18	Canonical, session-event, and workspace durable memory recall now share normalized FTS-backed retrieval with indexed `search_text` storage and Chinese tokenization support, but trust scoring, TTL policy, and richer derived-memory ranking are still limited
Documentation	A-	2026-03-13	Strong coverage across design docs, security, product sense, and quality tracking
CI/Enforcement	A	2026-03-13	8 CI workflows, convention-engineering (14 files, 11 checks), check:harness mirror gate
Contributor Experience	A-	2026-03-13	Clear tracks and recipes; could add more examples

Grading Criteria

A: Full test coverage, no known debt, documentation current, mechanical enforcement
B: Adequate coverage, minor debt tracked, docs mostly current
C: Coverage gaps, significant debt, stale or missing docs
D: Minimal coverage, blocking debt, docs unreliable
F: Untested, untracked, undocumented

Harness Maturity Assessment

Criterion	Status
Agent entry point (AGENTS.md)	Present, 102 lines, mirrored with CLAUDE.md
Architecture defined with enforcement	Present, DAG + boundary checks + CI
Progressive disclosure hierarchy	Present, 3-tier structure
Mechanical enforcement	8 CI workflows, convention-engineering (14 files, 11 content checks), check:harness, pre-commit
Quality tracked	This file
External context captured	Core beliefs principle #8 requires it