feat: critical file resilience framework — architecture, tests, and CI gates #732

Overview

Issue #730 exposed a class of bugs where squad.agent.md can silently disappear after upgrade/init. The immediate fix is tracked in #730/#731, but the root cause is systemic: the CLI lacks a principled approach to critical file integrity.

This issue tracks the systemic fix — a tiered resilience framework that prevents this class of bugs for all critical files, not just squad.agent.md.

Goals

1. Architecture principles — define which files are critical, what invariants they must satisfy, and how the CLI enforces those invariants
2. Test principles — establish test patterns that catch silent degradation before it ships
3. CI gates — automated checks that block PRs introducing silent-skip patterns

Linked Issues

• fix: squad.agent.md silently disappears after upgrade when template source is missing #730 — immediate bug: squad.agent.md silently disappears (root cause analysis)
• fix(#730): prevent squad.agent.md silent deletion — 3 code path fixes #731 — failing tests proving the bugs exist

Deliverable

A tiered PRD with concrete implementation specs for each layer:

• Tier 1: Immediate fixes (make fix(#730): prevent squad.agent.md silent deletion — 3 code path fixes #731 tests pass)
• Tier 2: Framework (critical file registry, validation, recovery)
• Tier 3: Prevention (CI gates, test patterns, lint rules)

See PRD in docs/proposals/critical-file-resilience.md (to be created).