feat: document advisor fallback to Agent tool when advisor() unavailable

[BaseInfinity]

Problem

The SDLC skill's Cross-Model Review section documents using codex exec for async review but doesn't document:

1. The advisor() built-in tool — exists in Claude Code but can be unavailable in some sessions
2. Fallback pattern — when advisor() returns "unavailable", the correct fallback is Agent(model="fable") (or whichever advisor model is configured)
3. The distinction — advisor() sends full conversation transcript automatically; Agent() requires a self-contained prompt. The fallback isn't free — you lose the auto-transcript and must brief the agent manually

What happened

Session had advisor() unavailable. Instead of spawning Agent(model="fable", prompt="review these recommendations...") as a fallback, Claude did a self-review — which defeats the purpose of cross-model adversarial review.

Suggested fix

Add a "Review Tool Availability" section to the SDLC skill that documents:

Review tool priority:
1. advisor() — sends full transcript, strongest review
2. Agent(model="fable") — manual briefing, still cross-model
3. Self-review — last resort, NOT equivalent to cross-model

And a note: "If advisor() returns unavailable, IMMEDIATELY fall back to Agent. Do not skip cross-model review."

Context

This is relevant for any project using the two-layer review model (Fable advisor + Codex safety net). The anticheat project's CLAUDE.md already documents "Spawned via Agent tool" as the Fable pattern, but the SDLC skill doesn't reinforce this.

[BaseInfinity]

Already addressed in skills/sdlc/SKILL.md line 109: **Advisor:** advisor() before plans; if down, spawn Fable subagent. Shipped in PR #412.