refactor: optimize config/doctor/status — PMRP->appendix, schema->bullets

[JFK]

Problem

commands/config.md (379), commands/doctor.md (471), and commands/status.md (192) have ~660 tokens of compressible prose, mostly dense procedures and pseudo-XML that read better as tables/bullets.

Scope (CTO/Explore review 2026-06-13)

• doctor.md: move the 85-line Plugin Metadata Resolution Procedure bash (~doctor.md:203-287) to an appendix/reference, keep a ~20-line summary inline.
• status.md: convert the review-block schema pseudo-XML <if/> branching (~status.md:48-97) to three bullet lists (v2 fields / v1 fields / synthesis rules).
• config.md: turn the 3-layer output-language policy (~config.md:13-26) into a 2x3 table; consolidate the repeated review.provider/review.model prose around its table (~config.md:49-70).

Acceptance criteria

• /config, /doctor, /status behavior unchanged (read-only commands — verify output still renders the same data).
• Token baseline (verdict parser: replace heuristic with structured ## Verdict: <token> contract #1) shows reduction.

Depends on

#1.

[JFK]

Closing after re-audit (see #96). Marginal value + real risk, no safe runtime-token win:

• doctor.md PMRP → appendix: a within-file appendix is still loaded whole, so it saves no runtime tokens. Worse, the PMRP's plugin-presence probe is load-bearing shared state — CONTRIBUTING.md:24/126 ("Presence checks belong in one place") requires it to stay in sync with start.md step 13b's ls -d ~/.claude/plugins/cache/... glob. Relocating it raises drift risk for ~0 token benefit.
• status.md pseudo-XML → bullets: convertible, but the <if/> nesting encodes the v1/v2 state-schema branching; flattening risks obscuring the backward-compat obligation for marginal gain.
• config.md 3-layer → table: the current bullets are already clean, and Layer C holds a load-bearing parser-token enum list a 2×3 table would mangle.

No safe, beneficial change here. Reopen if a specific low-risk cleanup is identified.