Lineage & Explainability Artefacts
For every output consumed downstream, produce a lineage artefact: what was retrieved, what tools ran, which model version, what policies evaluated.
This page is the complete instruction page for one pattern called "Lineage & Explainability Artefacts." It explains the problem this pattern solves, the idea behind the solution, when you should (and should not) use it, and what happens afterward — both the good effects and the costs.
This matters because building AI agents is not just about making them clever. It is also about making them safe and predictable. Following a well-tested pattern like this one helps avoid common mistakes, and shows you exactly which safety rules and regulations it connects to, listed under "Standards Mesh" on this page.
Context
Downstream consumers of agent outputs (reports, decisions, communications) need to understand provenance. Regulators require it. Without structured lineage, every question becomes an archaeological expedition.
Problem
Agent outputs without lineage cannot be trusted, audited, or safely re-used.
Forces
- Lineage completeness vs output size
- Machine-readability vs human-readability
Solution
Attach a lineage record to every material output: retrieval sources, tool invocations, model versions, prompt version, policy version, timestamps, agent identity, principal identity. Expose via an API; summarise for human readers.
Applicability
- Reporting agents
- Decision-support agents
- Content-generation for regulated use
Anti-Patterns
- Lineage buried in logs only
- Lineage stripped at export
Consequences
- +Trustworthy downstream consumption
- +Supports subject-rights requests
- −Artefact size overhead
- −Schema must evolve with the system