Runner metadata

2 min readReferenceUpdated 2026-05-19

What you'll find here

The stable metadata snapshot every adapter records at the boundary, so receipts and handoffs can explain which runner produced work — without leaking provider-specific fields into the stable contract surface.

Descriptive, not authorizing.

Receipt and handoff metadata is descriptive. It does not grant tool authority, prove that external execution occurred, or replace policy receipts for concrete side effects.

Stable keys

runner_id

runner_name

adapter

adapter_version

execution_mode

capabilities

trust_profile

capability_profile

policy_notes

Field

Storage

Where it lives

Receipt snapshot

capability_receipt

craik.capability_receipt.result.metadata.runner_metadata

Handoff snapshot

handoff

First unique runner snapshots on task receipts in craik.handoff.runner_metadata.

Runner-specific metadata

Runner-specific details remain nested under runner_specific. Preview adapters can preserve local fixture mode, live availability, executable names, or provider-specific session metadata without promoting those fields to the core contract.

Always redacted before storage.

Runner metadata snapshots are redacted before they are stored or copied into handoffs. Secret-like keys (tokens, passwords, API keys, credentials) are replaced with [REDACTED].

What's next

ReferenceRunner adapter contractHow adapters produce and preserve this metadata.GuideRunner preview workflowsSmoke-test that metadata appears in outputs.ReadRuntime contractsThe wider receipt and handoff shapes this metadata attaches to.