Repository navigation
Add shared policy contract roadmap with AgentFence #120
Description
Problem
agent-kernel and AgentFence both enforce tool-call policy, but at different integration points: agent-kernel in-process inside an agent runtime, and AgentFence as an external MCP/tool-call proxy. Users need to understand whether one policy intent can be authored once and enforced in both places.
Goal
Document and track a shared policy contract roadmap from the agent-kernel side.
Scope
- Map agent-kernel policy concepts to AgentFence policy language.
- Identify common concepts that should move into weaver-spec.
- Identify intentionally different semantics.
- Define shared fixtures for allow/deny/ask decisions.
- Clarify audit trace compatibility and differences.
Acceptance criteria
- agent-kernel docs include a shared-policy roadmap section or page.
- The docs link to the corresponding AgentFence roadmap/example.
- At least one portable policy intent is demonstrated.
- Follow-up implementation/conformance issues are identified.
Activity
dgenio commentedon Jun 13, 2026
Closing as a duplicate of #116.
This issue is one of five near-identical "Add shared policy contract roadmap with AgentFence" issues (#116, #117, #118, #119, #120) opened by the same author within ~3 minutes on 2026-06-07. #120's wording differs only trivially from #116 (same Problem/Goal/Scope/Acceptance intent). Consolidating all tracking onto the earliest one, #116, so the discussion isn't fragmented.
Note: #120 is referenced from several other issues (#131, #135, #138) as the shared-policy-contract tracking issue. Those references should be read as pointing at the surviving canonical issue #116 (and the related implementation issue #111). The closely related implementation issue #111 remains open; a maintainer may want to fold #116 into #111, but that consolidation is left open deliberately rather than assumed here.
Generated by Claude Code