A skills supply chain with the evidence regulated teams need.

Six areas cover a skill’s lifecycle, from packaging its instructions to tracing the decision it influenced.

The registry lifecycle

  • Publish & package

    Package open Agent Skills and plugins. Use a manifest to describe the intended behavior of each version.

    • Web, CLI, or API publishing
    • Inputs, outputs, required tools, and failure modes
    • Versioned, multi-file bundles
  • Curate & source

    Define where skills may come from and what needs review before use.

    • Approved upstreams and curated collections
    • Internal skills for private workspaces
    • Catalog discovery by domain
  • Secure & harden

    Review instructions, scripts, and declared access before installation.

    • Package scanning and isolated execution
    • A review queue with a record of decisions
    • A hardening report for changes made during review
  • Test & prove

    Collect behavioral evidence for each release and measure it against workspace expectations.

    • Golden-set, adversarial, and regression tests
    • Workspace-specific cases and thresholds
    • Re-testing when dependencies change
    Explore skill testing
  • Sign & distribute

    Bind approval to the version your agents actually install.

    • Content hashes and publisher attestations
    • Immutable version pinning
    • Catalog, CLI, and API access
  • Govern & audit

    Apply workspace trust policy and retain the evidence behind decisions.

    • Official, Verified Partner, and Community tiers
    • Release evidence and install history
    • Decision links through ElixirData

Where runtime controls belong

ElixirClaw owns runtime permissions, connectors, and decision boundaries. ElixirData owns the context and decision record. ElixirHub ties the release evidence to its source.