FOLD into 5-regen-verify (one class-kill, not a sibling): floor-gate src/v1/*.dag PARSE via shape-(b) — a non-#[ignore] parse-only witness over every src/v1/*.dag (cheap half of full_dsl_compiles, no heavy dsl/ compile). Closes 'src/v1 §7 authority gated only OUTSIDE the floor': src/v1 not in witnes

[briansrls]

Summary

Closes the gap: `src/v1/*.dag` files had no floor-level parse gate — the only existing gate (`full_dsl_compiles` in `src/v1/tests/src/pipeline.rs`) is `#[ignore]`-marked and never runs in CI. This lift is the cheap parse-only half of `full_dsl_compiles` ("shape-(b)"), folded into 5-regen-verify.

Files changed:

• `src/v1/stage0/src/bin/v1_src_dag_parse.rs` — new binary: `read_dir("src/v1")` walk (roster derived from the tree, not hardcoded), tokenize+parse each `.dag` via v1's own `v1_compiler_tokenize::tokenize` + `v1_compiler_parse::parse` (NOT serde_yaml), exit 0 on all-clean / exit 1 on parse error with typed diagnostic. Files are parsed in parallel threads (each gets its own `RcHashMap::new()` — no shared state; thread panics propagate via `std::thread::scope` — fail-closed). Includes `const SCAFFOLD_NOTE` (named dissolution trigger per §6 — not a comment).
• `src/v1/stage0/Cargo.toml` — `[[bin]] name = "v1_src_dag_parse"` entry (also marked SCAFFOLD in the TOML description).
• `dsl/tools/host_prelude.dag` — `ensure_v1_src_dag_parse_built()` alongside `ensure_gunbc_built()`.
• `dsl/tools/v1_dag_parse_transport.dag` — mirrors the existing `*_transport.dag` family (`dsl_compile_clean_transport.dag` pattern): `ShellProgram` built via `ensure_v1_src_dag_parse_built()` + `serialize_bash`, `run_v1_dag_parse() -> Bool` calls `shell.Exec.Run(script: serialize_bash(p: ...))`. NOT a new raw inline-bash fork.
• `dsl/test/claim/v1_dag_parse_witness_test.dag` — `test fn v1_dag_parse_witnesses()` auto-enrolled via `witness_discovery_scan_dirs = ["dsl/test/claim"]`; non-Rust, non-`#[ignore]`; runs CI floor batch-2 via `corpus_runnable()` (RunnableDiscoveryBatch).

Receipts (by-execution §5)

• Green (parallel): `./target/release/v1_src_dag_parse` → exit 0, "37 file(s) parse-clean", wall 1m35s (~2.4x vs serial 3m52s; floor-wall-neutral — fits within the 15min discovery corpus window).
• Red-on-injection (parallel): stray `#INJECTED_PARSE_ERROR` in `src/v1/00_core.dag` → exit 1, "parse error in 00_core.dag: expected item declaration (type, fn, func, service, resource, data, extern, pattern, interface)". Fail-closed survives parallelization. Reverted → exit 0.
• Disposition carrier slice-3 (core): ground std DeclarationRef String→record + mint std.decl_ref + emit-core rewrite #5632 branch verify (ff04439): checked out `Disposition carrier slice-3 (core): ground std DeclarationRef String→record + mint std.decl_ref + emit-core rewrite #5632`'s `src/v1/05_emit_rust.dag` changes into a scratch tree, ran against all 37 files → exit 0, "37 file(s) parse-clean". Merge-skew risk eliminated by execution; post-Disposition carrier slice-3 (core): ground std DeclarationRef String→record + mint std.decl_ref + emit-core rewrite #5632 constraint satisfied.
• Corpus compile: 1270 modules indexed, 0 new diagnostics (2 pre-existing layering-scan fixture intentional failures).

Follow-up noted (out of scope here)

The v1 seed parser has super-linear parse cost on some file contents: `05_emit_rust.dag` (479KB) = 98s single-threaded vs `04_infer.dag` (307KB) = 20s — not size-linear. Separate §1 seed-parser-perf finding; tracked for future attention, not this PR's scope.