What we shipped.

This release unifies the Helleze product and Hell skill into a single coherent system, updating documentation, specs, and source files across the board. It also introduces a safe start-fresh workflow for Hell, allowing projects to be reset without data loss risk.

New

• Safe start-fresh workflow added for Hell projectsA new reset flow lets you start a Hell project fresh without risking unintended data loss. The CLI, API, and skill reference have all been updated to support the reset workflow safely. ebb8953

Improved

• Helleze product and Hell skill unified into a single systemThe Helleze product definition and the Hell skill reference have been merged into one coherent structure, aligning documentation, specs, routes, models, and evals under a shared design. This makes the product and skill easier to reason about and maintain together. d124c8d

This release is dominated by two large internal builds: the scout SEO research pipeline (stages 1–8, now in src/lib/scout/) and the artifact/judgment architecture (Phases 1–6 plus the synthesis outbox and Project Intelligence). The scout library goes from a proof-of-concept to a fully typechecked, tested module that discovers competitors, scores SERP beatability, maps niche channels, and traces rival growth — all without hardcoded lists. The artifact architecture adds durable schemas for assets, galleries, brand kits, message libraries, product observations, voice constitution rules, artifact adapters, and result tracking, wired together through a shared feedback and synthesis engine. Supporting these are extensive planning documents, SEO strategy research, an OG image composition trial, and a Google Search Console integration.

New

• Agent-native asset engine, galleries, and shared judgment memory built (Phases 1–4)The core artifact architecture was implemented across four phases: Phase 1 adds immutable artifact revisions with a shared ArtifactFeedbackEnvelope write-through; Phase 2 adds a deterministic artifact_index projection across draft, OG, and gallery history; Phase 3 adds a project-isolated message library (category, promise, headline, CTA, objection/response, customer language) with versioned lifecycle; Phase 4 adds product observations, Voice Constitution rules, and a generalized synthesis outbox. Migrations 0008–0016 cover assets, brand kits, project judgments, galleries, artifact feedback, provenance, artifact index, message library, and product heart. 3fb306d 7d1e434 2bf0a8d e183032 aa84d47
• Artifact adapters and result tracking built (Phases 5–6) with durable synthesis outbox drainPhase 5 adds a typed adapter registry (src/artifacts/) so each artifact type owns only its schema, renderer, destination rules, and validation — the shared feedback and memory engine is inherited. Phase 6 adds artifact results, outcome ingestion, and a publish queue. A durable synthesis outbox drain (synthesis-scheduler.ts, migration 0019) ensures synthesis proposals survive worker restarts. Migrations 0017–0019 cover artifact adapters, results, and the outbox drain. f56771c e9a5c19 d21c802 3003ee2
• Project Intelligence bootstrap and research stages built, including Google Search Console integrationA project_intelligence system (src/lib/project-intelligence/, migration 0020) bootstraps a structured intelligence pack for any project: competitor trails, engagement classification, backlink pages, public results, query suggestions, roundup competitors, search demand, search intent, and venue validation. A Google Search Console integration (src/lib/gsc/, migration 0021) adds OAuth-free service-account access to GSC performance data and wires it into the research stages pipeline. c1132f2 0b17d55 84835c9