Adapter capability matrix

Published adapter capability matrix (goal.md §13.1). Levels: A0 package/version detection, A1 build/typecheck/test observation, A2 static symbol resolution, A3 runtime symbol instrumentation, A4 clean Sample + Contract verification. No adapter pretends a level it does not have.

How to read this table

Level	What it means
A0	Package and version detected from the lockfile
A1	Build, typecheck and test results observed
A2	Static symbol resolution — which API was used
A3	Runtime instrumentation — the call itself observed
A4	Clean sample verified against a contract

Symbol confidence	What it means
EXACT	Resolved with full type information
PROBABLE	Resolved from imports and call sites
UNKNOWN	Dynamic usage — package-level evidence only

Ecosystem	Adapter	Package managers	A0	A1	A2	A3	A4	Symbol confidence	Notes
npm	node-typescript@1.0.0	npm, pnpm, yarn	✓A0 — Package and version detected from the lockfile (supported)	✓A1 — Build, typecheck and test results observed (supported)	✓A2 — Static symbol resolution — which API was used (supported)	—A3 — Runtime instrumentation — the call itself observed (not supported)	✓A4 — Clean sample verified against a contract (supported)	PROBABLE	Reference ecosystem. Lockfile-resolved versions (package-lock v2/v3, pnpm-lock v9, yarn classic). Static import/member analysis without a type checker yields PROBABLE symbols; EXACT reserved for future TS type-info integration. Verifier adapter node-typescript@1 supports clean resolve/compile/contract.
pypi	python@1.0.0	pip, uv	✓A0 — Package and version detected from the lockfile (supported)	✓A1 — Build, typecheck and test results observed (supported)	✓A2 — Static symbol resolution — which API was used (supported)	—A3 — Runtime instrumentation — the call itself observed (not supported)	—A4 — Clean sample verified against a contract (not supported)	PROBABLE	Resolved versions from uv.lock, pinned requirements, poetry.lock (best-effort). AST-free regex import analysis; dynamic attribute access degrades to UNKNOWN. Sample verification uses the shared container verifier only when a contract is expressible without native builds.
golang	go@1.0.0	go	✓A0 — Package and version detected from the lockfile (supported)	✓A1 — Build, typecheck and test results observed (supported)	✓A2 — Static symbol resolution — which API was used (supported)	—A3 — Runtime instrumentation — the call itself observed (not supported)	—A4 — Clean sample verified against a contract (not supported)	PROBABLE	go.mod/go.sum resolved versions; go/parser import + selector analysis. EXACT requires go/types integration, not claimed in v1.
cargo	rust@1.0.0	cargo	✓A0 — Package and version detected from the lockfile (supported)	✓A1 — Build, typecheck and test results observed (supported)	✓A2 — Static symbol resolution — which API was used (supported)	—A3 — Runtime instrumentation — the call itself observed (not supported)	—A4 — Clean sample verified against a contract (not supported)	PROBABLE	Cargo.lock resolved versions; use-statement analysis. Macro-expanded and dynamic usage degrades to UNKNOWN (goal.md §13.5 conservative start).

A3 is not provided by any Public v1 adapter (goal.md §19).