Product experience
Owns the public site, controls, schemas, APIs, review flows, and deployment-ready application experience.
Reuse discipline
Build AgentFoundry from proven engineering patterns without exposing tool sprawl.
The product owns the user-facing experience. Engineering references are translated into first-party contracts, routes, schemas, and safe boundaries.
Execution foundation
Provides governed execution, coordination, and verification ideas translated into first-party contracts.
Quality foundation
Provides the verification, policy, and evidence ideas translated into product-level quality controls.
System coordination
Keeps deployment, governance, and review decisions aligned without diluting the product boundary.
Every public claim is reduced to a product-safe abstraction.
AgentFoundry publicly describes product capabilities, control layers, evidence flows, and deployment boundaries without exposing internal paths, private source indexes, confidential repo structures, or implementation-specific execution inventories.
Product shell
User-facing patterns become a first-party product experience rather than raw implementation notes.
Execution model
Run coordination, approval state, and evidence capture are translated into controlled product contracts.
Verification model
Checks, reviews, risks, and safe-stop posture are preserved as product-visible trust layers.
Reuse lanes
Translate proven patterns into product capabilities.
Product UXworkspace patterns → turn into the AgentFoundry product shell, review flows, intake, and engineering console
App patternsproduct patterns → turn repo coordination, work sessions, coding execution, and review handoff into AgentFoundry product capabilities
Execution machineryexecution foundations → translate proven execution behavior into coding-agent records, policy checks, and safe-stop evidence
Tool, evaluation, and sandbox foundationtool, evaluation, and sandbox layers → show them as tool permissions, governance, and Evidence Reports instead of raw tools
Engineering template and app references
Existing builder and app patterns become examples for engineering-agent templates and review flows.
Execution and tool references
AgentFoundry hides tool complexity and gives teams one controlled way to run, inspect, and approve the work.
Reliability references
Run report, evaluation, policy, and quality material stays internal until real engineering traces justify hardening or scale-up.
Rules for safe use
!Reuse proven foundations when they can become clear product capabilities.
!Keep implementation details private; explain work records, execution controls, tool permissions, governance, and Evidence Reports.
!Keep the public site tied to product behavior; every claim should map to how work is submitted, run, checked, and reviewed.
!Run AI coding work only with evidence, approval points, policy decisions, risk notes, changed-file records, and safe-stop notes.