From draft to install, every gate leaves evidence.

Follow a skill from its first manifest to a pinned release that a workspace can install. Each gate records what was checked and who approved it.

The five gates

  1. Draft

    The publisher builds the skill and declares its tools and inputs in the manifest.

    • Evidence: Manifest, declared tools, and publisher
  2. Scanned & tested

    Review the package and run sandbox, functional, adversarial, and regression tests.

    • Evidence: Scan findings, sandbox trace, test results, and regression comparison
    See how skills are tested
  3. Signed

    Bind the publisher key and attestations to the release’s content hash.

    • Evidence: Content hash, signature, attestations, and approver
  4. Published

    Make the immutable version discoverable by trust tier and domain.

    • Evidence: Version, trust tier, publish date, and changelog
  5. Installed

    Check that the release’s tier meets workspace policy.

    • Evidence: Workspace, installer, pinned version, and agent bindings

Who owns each control?

The registry governs the skill. Runtime permissions and decision records belong to the rest of the Elixir Platform.

Scroll horizontally to see the full comparison.

ControlOwnerQuestion it answers
Publishing, scans, evaluations, and signaturesElixirHubWhat does this version do, and who approved it?
Discovery and workspace trust policyElixirHubMay this workspace install the release?
Runtime permissions and tool accessElixirClawWhat may the agent do right now?
Context Graph and decision recordsElixirDataWhat was decided, with which skill version, and why?

Install where agents already work

  • ElixirClaw agents

    Attach a reviewed release to an enterprise agent with its version pinned.

    Explore ElixirClaw
  • Coding agents

    Browse reusable skills for supported development tools.

    Browse the catalog
  • CLI and API

    Use the registry’s publishing and discovery interfaces in your workflow.

    View API documentation