fix(cli): render **bold** in docs topics instead of printing the markers (#7277) - #7281

Reviewed at aaf4a0217d8b4f523903fc0e1338f636135ed353.

The renderer now handles the reported bold-around-inline-code form, and the regression renders the actual compositions topic (packages/cli/src/commands/docs.test.ts:68). It reuses the existing color helpers, including their non-TTY behavior. Of the three submissions, this most directly pins the original report; #7327 has equivalent formatting, while #7328 still leaves the original nested markers.

Nit: packages/cli/src/commands/docs.ts:72-77 processes bold after removing backticks, so literal ** inside a code span can also be consumed (for example, the code span **not prose**). Preserve code-span contents while retaining the test for bold around code. This formatting edge does not block the reported topic fix.

Validation at this head: bun run --cwd packages/cli test src/commands/docs.test.ts → 4 passed. Built dependencies were reused from the preceding CLI build. Fresh independent source review found no Important defect; no live browser or native Windows run. No prior substantive reviews or inline comments were present.

Need: yes, requested starter issue · Size: ok · Reuse: existing CLI policy, color and checker helpers · Simpler: already minimal · Structure: fine
Risk: low — bounded CLI output/checker change

— hf-review

Verdict: APPROVE
Reasoning: The requested behavior is covered by focused tests and reuses the existing contract. No Important blocker found.

Reviewed at a49c2baa2e9033a4312cda223f5d395f4320f895.

The prior literal-asterisk nit is resolved: packages/cli/src/commands/docs.ts:72 now selects code spans before interpreting bold, and the bold alternative treats nested code as opaque. The original compositions-topic regression remains covered, and the new case preserves literal ** both alone and inside bold (packages/cli/src/commands/docs.test.ts:68). No new blocker found.

Validation: independently extracted the actual formatInline and renderMarkdown source, transpiled with Bun and executed in a VM. 21 assertions passed across identity/non-TTY and ANSI color adapters: the actual compositions topic, original bold-around-code, literal asterisks in nested code, adjacent spans and unmatched markers. This probes formatter behavior; it does not load the full CLI module, actual picocolors, or run the repository Vitest suite. Full current command and test files read; delta is limited to the formatter and its regression.

Need: yes, starter issue and review feedback · Size: ok · Reuse: existing color helpers · Simpler: bounded single-pass token choice · Structure: fine
Risk: low — isolated CLI documentation formatting

— hf-review

Verdict: APPROVE
Reasoning: The reported topic formatting remains correct, and the follow-up preserves code contents without a new dependency or broader renderer changes.