tier-escalatelisted

# tier-escalate Reads every `SKILL.md` under your `.claude/skills/`, classifies each skill's LLM-bearing roles with a fixed taxonomy, then applies the up-direction escalation rule to decide which (and only which) skills contain a phase worth escalating a *session* to **Fable 5** (`claude-fable-5`, one tier above the Opus default). It emits one artifact: a human-readable **Fable escalation map** (markdown, grouped by verdict) that tells the operator exactly when `/model claude-fable-5` buys quality and when it only multiplies cost. This is the UP-direction sibling of `tier-offload` (which finds the slices safe to route *down* to a cheap local model). It is **discovery + map only** — it identifies the escalation-worthy phases and their triggers; actually switching a session to Fable stays a per-session operator action afterward. It never edits a skill or CLAUDE.md. ## When to use - You want to know which of *your own* skills have a phase where escalating the session to Fable 5 buys real quality — and which would just double cost for the same output. - You want to regenerate the escalation map after adding or changing skills (the escalation surface drifts as skills evolve). - You want the CLAUDE.md model-paragraph escalation list regenerated from the current skill set as a paste-ready block, instead of hand-maintaining it and letting it drift. ## Background — the up verdict rule (the Fable signature) The escalation's one governing rule (the operator's tiering policy, encod