TuxRacer conservative Tcl lexical repair

Base: 5c8f6750b52d320130da300cbba2336dc0fdc53c. Production SHA-256: 69257615178adda7568036fc49e2f20b4ca602c6d3b734f1238caf6143021b17.

The unchanged test/tuxracer-parent-relative-independent-oct28.py first ran against its original pinned production source into fresh owned fixtures. Its parent boundary passed, its parent-stack mutant was killed, then inert braced set data falsely raised a missing literal source for ../common/absent.tcl. The original test and its rejected-source pin remain byte-identical.

The scanner now recognizes top-level commands while retaining multiline braces and quotes as word data, comments at command boundaries, escaped characters, and ungrouped newline/semicolon boundaries. Control bodies are not evaluated. Bracket substitution, malformed/incomplete grouping, computed source arguments and source continuations remain UNKNOWN. Unsupported bracket syntax ends the remaining file audit conservatively. Non-source commands are explicitly UNKNOWN. This is a lexical subset, not complete Tcl parsing or runtime resolution.

Literal target normalization and confined root/script-directory candidate checks are unchanged. Missing real literals still reject; all-escaping targets remain UNKNOWN. Runtime CWD, substitution and filesystem/runtime resolution remain UNKNOWN. No Tcl execution, original-media extraction or runtime qualification was performed.

literal-sources DIRECTORY INVENTORY exposes the same lexical API as a production CLI for local static audits; it reads the supplied script inventory and does not execute Tcl. Like the existing API, inputs are trusted audit inputs.

Validation on the production hash above:

• The transparent test/tuxracer-lexical-independent-adapter-oct29.py replaces only the original harness's source-hash assertion in memory. It runs the exact original fixtures and behavioral assertions successfully. It does not claim that the old pinned harness runs unchanged against a changed source hash.
• test/tuxracer-lexical-commands-oct29.py passes 23 cases through both the API and actual production CLI: inert braces/quotes, nesting, comments, escaped newline/semicolon/brace, real source after data/comments, missing literals, parent candidates, encoding option, root escape, computed/control syntax and malformed grouping. Restoring the old boundary matcher is killed by the exact inert braced-data regression.
• Reuse the unchanged eight parent/static closure checks through predecessor receipt scratch/usage-watch/tuxracer-parent-relative-independent-accept-oct28/receipt.json (SHA-256 1a4a2d59f4d0c3fb7c0f08113070fca50ed082fbe5701df0692c5824b2cb4fd2) and coordinator review (SHA-256 78753c972f8488682251674553a0a3b16bc3cd4a64f3a30bc8f183ca335d4628). The inherited original manifest remains UNSEALED; no repeated payload census.

Owned evidence: /tmp/tuxracer-cwd-contained-lexical-repair-oct29/receipt.json. Resource allocation remains held through coordinator publication/closeout. Fresh independent acceptance is required before serialized main integration. No main push or public deployment is authorized by this task.