Lineage v1.0.0

The first V1 distribution release: install, publish, pull, inspect, enable, and run a package, end to end, without cloning the repo.

What V1 supports

• Install without cloning: curl -fsSL https://agenticlineage.vercel.app/install.sh | sh (primary path), or go install github.com/agentic-lineage/lineage/cmd/lineage@latest for Go developers.
• Author and publish: lineage login (GitHub OAuth device flow, no admin-minted tokens), lineage package publish <path> against the live registry, with content-addressed digest verification and interruption-safe (draft-then-finalize) uploads.
• Receive and enable, one command: lineage add <ref-or-local-archive> [--yes] - accepts a registry ref or a local .tgz from lineage package export, fetches/imports, shows exactly what it contains (skills, workflows, agents, policies, declared capabilities) and asks permission before enabling. Re-running against an already-present package is a no-op, not an error.
• Non-coder receiver path: a copy-paste prompt (docs/bootstrap-prompt.md, embedded per-package on each /packages/<name> page) that a fresh Claude/Codex session can run unattended - install the CLI, fetch, inspect, enable - reviewed and tested for the prompt-injection boundary (see that file's Verification status section for exactly what was and wasn't tried).
• Package setup playbook: a package can declare tracker files or directories its workflow expects; the receiver sees the plan and approves it before anything is created, and re-running is idempotent.
• Integrity and safety active throughout: secret scanning, path-traversal guards, and independent digest verification run on every publish, pull, and import - the registry's own metadata is never trusted blindly.
• Provider and capability disclosure: /packages and /packages/<name> show which providers (claude/codex) a package declares an entrypoint for and what filesystem/network access it asks for, searchable by provider; the same data is available via lineage inspect --yaml / lineage package validate --yaml for scripting.
• Documentation: README and the website both show the real, current install → publish → pull → enable flow.

What is experimental

• The bootstrap prompt's prompt-injection defenses are verified against one real adversarial test package - a real, useful check, but not a formal security guarantee. Wording changes should be re-verified.
• Package setup covers files and directories only, not simple local databases.

What is explicitly out of scope (non-goals)

• No marketplace, billing, enterprise UI, vector database, or cloud execution - none of this is part of V1 or any currently planned release.
• No enforcement of declared capabilities at runtime - they remain purely declarative (ADR 0006), shown to the receiver before enabling, not sandboxed.
• No fully type-enforced shared state machine across every artifact-processing code path - the CLI behavior is unified (local and pulled artifacts converge on the same inspect/confirm/enable pipeline), but deeper code-level unification remains a documented follow-up.
• No JSON output mode - structured output is YAML (--yaml), by deliberate choice, consistent with the rest of the project's manifests and config.

Compatibility

No breaking changes for already-published packages: releases published before this version simply show provider/capability fields as "not declared" rather than erroring.

Verification

Built from master commit cba66f1380e05fe3c329f6e202a60c3c0962eaa3. Every change in this release was individually reviewed and tested in its own PR before being promoted; see PR #126 for the full list and this release's own verification notes.

Install artifacts

lineage-darwin-amd64, lineage-darwin-arm64, lineage-linux-amd64, lineage-linux-arm64, lineage-windows-amd64.exe, with checksums.txt for verification.