docs(demos): offer sandbox placement for /promote, now that --exclude-prompt exists

[pdettori]

Summary

Completes the follow-up #218 promised and #217 could not take.

#218 is merged and this branch is rebased onto it, so the diff below is this change alone.

#217 had to install /promote into real user scope, and justified it with a constraint that no
longer holds — that a command living in the project it promotes would ship itself into every bundle.
--exclude-prompt removes that, so the placement that actually buys local/remote parity becomes
available.

Both placements, and what each costs

	Option A — real user scope	Option B — in the sandbox
setup	cp into ~/.claude/commands/	cp into the sandbox, launch HOME=$SANDBOX claude
local agent sees	your whole ~/.claude	exactly what the promoted run sees
cost	"it behaved the same locally" is weaker evidence than it looks	a re-auth, and your own skills for the duration

Option B is the more honest way to perform the demo, since parity is its whole claim. Option A stays
the easy path for following along.

/promote self-excludes rather than making the reader remember

A Context probe reports whether .claude/commands/promote.md exists in the project, and the body adds
--exclude-prompt promote only in that case. The condition matters as much as the flag: passing it
unconditionally would trip the flag's own typo guard (prompt_exclude_unmatched), which exists
precisely because an unmatched exclusion ships the prompt you meant to omit.

The reassuring part, measured

Both options produce the same bundle, so every digest quoted in the walkthrough holds for either:

Option B, no exclusion  → 19456 bytes   (the command itself travelling)
Option B, self-excluded → 12288 bytes   sha256:43b8c4c0…   ← byte-identical to Option A

Also reconciles what the rebase exposed

README.md carried #217's claim that a project-local /promote "would ship itself into every bundle"
two paragraphs above #218 documenting the flag that prevents it, plus a doubled lead-in from the two
edits landing separately. Both fixed.

The "Notes and limits" fidelity entry is rewritten rather than deleted: Option A's gap is real, it
is now a choice rather than a constraint, and whoever performs this should say which option they ran if
asked whether local matched remote.

Verification

• make lint (9 hooks) and make test-deploy (129 checks) pass.
• The demo is 14 passed / 0 failed against kind.
• Option B exercised end to end from a scratch sandbox, with and without the self-exclusion, to
  produce the two byte counts above.
• The command test gained three checks pinning the self-exclusion, including that it is conditional.

Assisted-By: Claude Code

[pdettori]

Rebased onto main now that #218 is in (e1fc6df). The diff is a single commit touching 4 files — #218's commit is gone from it — and the stacking caveat is out of the description.

Re-verified against the new base, not just carried over: make lint (9 hooks), make test-deploy (129 checks), promote.test.ts 32/32, build.test.ts 19/19, and the demo at 14 passed / 0 failed against kind.