TITAN v7.2.1 — the full Fable-5 discipline, one switch

TITAN v7.2.1 — Reliability Mode composes all four Fable-5 disciplines

Building on v7.2.0 "Conscience", one switch (agent.reliabilityMode: true) now
gives a local model the complete disciplined-process nervous system of a
frontier agent:

1. Plan-first — on multi-step tasks, write a numbered plan and verify each step
2. Verify — never claim a side-effect without the tool that did it (the wall)
3. Self-critique — review your own draft adversarially, append honest caveats
4. Escalate — confirm before destructive / public / spend / shared-service actions

Live-proven: a single "refactor X, delete the backups, deploy to prod" request
triggered all four at once — the model investigated first, hit the approval gate
on the destructive step, and flagged its own unverified assumptions.

This is how a local model reaches Fable-5-class reliability — the discipline
is enforced by the harness, on whatever brain you plug in.

Off by default. Suite: 375 files / 8,180 passing / 0 failing.