v1.55.5

📦 Standalone skill ZIPs (claude.ai, Claude Desktop, Cowork, any Agent-Skills assistant): https://github.com/bjcoombs/ai-native-toolkit/releases/tag/standalone-skills-v1.55.5

What's Changed

Features

• feat: SVG accessibility metadata and golden-render colour tests by @bjcoombs in #243
• feat: Pin lizard + schema-aware semver diff reliability; infra-safe assess gate by @bjcoombs in #244
• feat: Emit override_contradicts_signals finding and disclose config-suppressed findings by @bjcoombs in #245
• feat: Reconcile run-context invariants at finalize with run_id stamps and Layer 6 cap by @bjcoombs in #246
• feat: /assess consent lifecycle - decline provenance, consent batching, uninstall path by @bjcoombs in #247
• feat: wiki-integrity guardrails - orphan hotspot pruning + log.md integrity chain by @bjcoombs in #249
• feat: Ship deterministic badge by default, funnel LLM grade to report by @bjcoombs in #250
• feat: actions.json v2 schema with status lifecycle and derived mode by @bjcoombs in #248
• feat: scope /assess to a subtree with a path argument by @bjcoombs in #251
• feat: Lead report with deterministic trustworthy content above the fold by @bjcoombs in #252

Fixes

• fix: Layer-scorer untrusted-data guard and author-time doc staleness by @bjcoombs in #242

Other Changes

• test: add self-dogfood invariant check for /assess deterministic output by @bjcoombs in #253

Full Changelog: v1.53.3...v1.55.5