v5.0.0

v5.0.0 Tiers 4 & 5 — Capability Introspection + Multi-step Compliance

Delivered: Extends the graduated proof model from Tiers 1–3 to Tiers 4
(Capability Introspection) and 5 (Multi-step Compliance Chain), giving
defenders deeper verifiable evidence of AI agent compliance — still
with no secrets transmitted, fully backward-compatible with v1.0–v4.0.

Key accomplishments:
- Tier 4 payloads: base64-encoded sorted tool/capability lists via /cb/v4/
- Tier 5 payloads: deterministic seed in JSON-LD, server-verified 3-digit
  proof ((seed+a)*b) %mod via /cb/v5/
- Additive SQLite migration (t4_capability, t5_proof, t5_proof_valid)
- Monitor TUI EVIDENCE column + always-visible detail pane + 6-state
  filter cycle + always-show chrome policy
- Markdown report Evidence column + exec summary T4/T5 rows
- test-agent 5-tier scorecard (n/5 score, exit codes 0/1/2 preserved)
- README Proof Levels with concrete examples, Ethics T4/T5 callouts,
  Project Status through Phase 15, TODOS.md ## Shipped section
- Landing page live stats extended with T4/T5 counts (|| 0 fallback)

Cargo.toml version synced to 5.0.0; binary self-reports honeyprompt 5.0.0.

Single-day execution (2026-04-24 → 2026-04-25), 3 phases, 10 plans,
+2806/-87 Rust LOC, 212/212 tests passing.

See .planning/milestones/v5.0-ROADMAP.md for full details.