Make work resumable
Keep the goal, scope, work items, completion conditions, and next responsibility explicit.
Follow the workflow →Shared across work
Common foundation
Keep AI work traceable from a bounded request to the next decision. The common foundation connects work progression, stable references, execution records, verification, and handoff. Individual Skills supply the method for each particular job.
The shared operating foundation
What every run can build on
These responsibilities connect the work. Each shared contract defines when its controls apply.
Use the right definitions
Resolve the selected Skills and Knowledge through stable XIDs, and preserve which references were used.
Understand stable references →Keep the basis visible
Record outputs, rationale, checks, and unresolved items. Report what matters for the next decision.
Explore records and reporting →Check and hand off
Verify the work record, review output quality when required, and transfer the result with its remaining concerns. People retain acceptance authority.
Understand handoff and acceptance →Available now and in development
Published capabilities and ongoing repository work are distinguished below.
Available in the published package
A traceable foundation for AI work
Use stable XIDs to resolve Skills and Knowledge, record execution progress and evidence, and check the work record before handoff. The Skill Run Dashboard makes recorded state visible for human review.
Common foundation · current development · quality
Preserve accepted output quality across runs
The common quality protocol connects per-run review and optional model qualification to target-specific Knowledge profiles. Those profiles retain approved formats, examples, inputs, and review criteria; each Skill defines its output requirements and references the applicable Knowledge. A format check alone does not establish content quality; human acceptance remains explicit.
Common foundation · current development · planning
Connect plans to execution evidence
Local plan views bring tasks, dependencies, completion conditions, confirmation questions, and evidence together. Dashboard views and generated Markdown and Mermaid views help people inspect progress. These are shared record and display tools; producing a plan remains the responsibility of the selected Skill or work instruction. Task completion and acceptance of the business outcome remain separate decisions.
Common foundation · current development · integration trial
Share selected work state with Azure DevOps
The current trial includes workspace-scoped, test-only connection checks and a reviewed mapping design for local and Azure work items. The next integration work prepares selected Task state and result summaries while keeping artifacts authoritative locally. This integration shares execution state; individual Skills still own their work methods. Full synchronization and production writes are not available in this trial.
Common foundation · reading guide
Choose by purpose
Start from the canonical model, then read only the depth required for the current question.
Understand the need and the model
See which AI characteristics require an operating structure and learn the canonical elements.
Operate AI work
Understand how bounded responsibilities connect and where tool, AI, and human decisions belong.
Implement, improve, and distribute
Connect the repository structure to evidence-led improvement and distribution of the next version.
Common foundation · concepts
Canonical concepts
Permanent concept pages define the problems XRefKit is intended to solve and connect them to the operating model.
→
Permanent problem definition
Judgment Replication
Organizations do not scale by adding processing capacity alone. They scale by replicating judgment—and traditional organizations do this slowly, lossily, and repeatedly. XRefKit replaces judgment retransmission with versioned distribution of stable references. Package and Core releases carry versions; XID references do not pin a version or include history.
Optional client connection
Connect through MCP when useful
The common foundation can be used without MCP. MCP is an optional discovery and delivery interface for bringing managed Skills, Knowledge, and contracts to AI clients.
What MCP shares
An organization maintains Skills, Knowledge, execution contracts, catalogs, and XIDs as managed XRefKit assets. MCP presents the catalogs, then selects and resolves only the assets needed for the current purpose. MCP resolves stable XIDs against the current managed content available to the configured provider, without loading the entire repository into every AI context. Versioning belongs to the distributed package or Core release, not to the XID reference; a Skill does not embed a version or history in that reference.
MCP is a read boundary for discovery, selection, resolution, and delivery. Skill execution, client-side tool side effects, Evidence recording, and closure remain on each user's client. People can therefore share the same Knowledge and procedures without merging their execution permissions or work records.
MCP shares managed definitions and Knowledge. The MCP server does not perform business work on behalf of its users.