refactor(audit): share the flag-presence verdict and tighten the line classifier

[brettdavies]

Summary

A cleanup pass over the stack below it (#163 to #179), with no change to any audit result.

Audits:

• Seven audits ended in the same shape: pass when the help declares one of the names, otherwise warn with the message and the dash-rule note. flag_presence::pass_or_warn holds it once, for p6-color-flag, p2-raw-flag, p7-verbose, p2-json-aliases, p7-quiet, p8-install-all and p1-flag-existence.
• p1-flag-existence and p1-non-interactive each kept a copy of the help-on-bare markers and of the sentence that says no non-interactive flag is declared. non_interactive.rs owns both.
• The declared [p2] json_probe call, which p2-json-output and p2-schema-print both run, moves to declared_probe.rs. json_output.rs drops from 239 to 188 lines of non-test code.
• subcommands_to_probe takes the help it always had, in place of an Option.

Flag model:

• The classifier kept two fields that were never set together: the description column, and the indent of a definition still waiting for a description. They are one Description enum.
• Lead had three variants and one question asked of it. It is a predicate, and the bracket check sits where it is used.
• column_of moves to pieces.rs, so classify.rs and header.rs no longer import each other.
• Piece kept two join flags that were only read together. They are one.
• The tokenizer's TAB handling is removed: normalize expands TABs before any line reaches it.

Tests read help fixtures through one fixture() helper.

Changelog

No user-facing changes.

Type of Change

• refactor: Code refactoring (no functional changes)

Related Issues/Stories

• Story: n/a
• Issue: n/a
• Architecture: docs/plans/2026-10-06-0034-fix-help-flag-names-across-frameworks-plan.md (the simplification pass after U10)
• Related PRs: fix(audit): pass p8-install-all only when skill install declares --all #179 and the stack below it (test(audit): pin what the flag parser reads from a capture of every help layout #163 to fix(audit): detect an output flag from definitions and probe it as spelled #178)

Testing

• Unit tests added/updated
• Integration tests added/updated
• Manual testing completed
• All tests passing

Test Summary:

• 1315 passing, the same tests as the base apart from one tokenizer table row (a TAB-separated description, which normalize owns) and one assertion on the removed None argument.
• Snapshots: none change. Every one of the 71 fixture captures parses to the same definitions.
• The full registry capture set (910 help texts) was rendered through the base and head parsers: the two renderings are identical, file for file.
• Self-audit: cargo test --test dogfood passes.

Expected moves. Written before the corpus run.

None. Replaying the full registry capture set through the base (#179) and head builds moves no row.

Corpus before/after. Run after the table above was written. Base 8a8ae3ef2419, head 8134d09b7510. Image sha256:05db16cf226cd14a93a2682aea41b72d3e6b4d240cadba006f2881d3ffa996b3, network bridge. 98 tools selected, 95 scored under both builds, 1 rerun three times (mods).

Moved rows (0)

None.

Derived fields moved (0)

None.

No row and no derived field moved, as expected. mods p3-should-about-long-about, the one row on the A/A noise list, did not move. No row is unstable, no harness bug is flagged, and no tool failed to score under one build only. cursor, nvidia-smi and xai-grok-build are absent from the image and ran under neither build.

Files Modified

Modified:

• src/audits/behavioral/: color_flag.rs, raw_flag.rs, verbose_flag.rs, json_aliases.rs, quiet.rs, install_all.rs, flag_existence.rs, non_interactive.rs, json_output.rs, schema_print.rs, force_yes.rs (test helper only), mod.rs
• src/runner/help_probe/flags/: classify.rs, header.rs, pieces.rs, mod.rs (test helper only)
• src/runner/help_probe/fixture_snapshots.rs, src/runner/help_probe/mod.rs: the shared fixture() reader

Created:

• src/audits/behavioral/flag_presence.rs
• src/audits/behavioral/declared_probe.rs

Renamed:

• None.

Deleted:

• None.

Breaking Changes

• No breaking changes

Deployment Notes

• No special deployment steps required

Checklist

• Code follows project conventions and style guidelines
• Commit messages follow Conventional Commits
• Self-review of code completed
• Tests added/updated and passing
• No new warnings or errors introduced
• Changes are backward compatible (or breaking changes documented)