autogen · Issues· 1074 open

Proposal: GuardrailProvider protocol for tool call interception

HandoffMessage.context splices peer-authored messages into another agent's model context with their declared LLM role intact

AssistantAgent may ask for approval before executing a tool

Proposal: Capability-scoped tool authorization for AutoGen multi-agent pipelines

源码里的非致命小问题

Fix stale documentation links and redirect fallback text

CodeExecutorAgent defaults execute any participant's code blocks on the host without approval; documented dangerous-command sanitizer does not exist

ChatCompletionCache omits tool_choice from the cache key, so 'must call a tool' and 'must not' share one entry

Prevent Margin Leaks & Surprise LLM Bills: Add a Hybrid Spend Firewall

autogen-ext: mcp 2.0.0 breaks MCP tools (no upper bound, uses removed/renamed 1.x APIs)

Paid security-scan MCP integration (autogen + ScanPay x402)

TextCanvas.apply_patch accepts stale hunks without validating context lines

Deterministic Sub-50µs Pre-Flight Invariant Gating for UserProxyAgent Code Execution

Support Approval Func in BaseTool in AgentChat

Feature: Agent trust verification via MoltBridge for cross-org multi-agent scenarios