- Document that the framework always prepends DEFAULT_SYSTEM_PROMPT before the
component systemPrompt(), and its behavioral implications.
- Correct §7.6: skillBox.codeExecution() is enabled and bound to the conversation
workspace; add the related security note in §7.6 and §10.
- Update §7.4 for container-based DI of skill tools (no-arg ctor is only a
fallback) and native YAML inline-list tools; drop the dead SkillToolManifest
reference and add a troubleshooting row for instantiation failure.
- Add §5.6 covering the AgentSessionFactory SPI for custom persistence backends.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add ChatUsageTrackingHook that accumulates ChatUsage from every
PostReasoningEvent within a single process() call, expose it via
ReActAgentContext#getChatUsage(), and emit a per-step usage line in
ReActLoggingHook. The hook is cached on AgentSession and reset at the
start of each process() so the snapshot reflects the full invocation
(not just the last reasoning step).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add notes on framework integration prerequisites, event listener
threading, build-time capability methods, session reuse semantics,
built-in tool risks, and two new troubleshooting entries.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Introduce FlowEvent/FlowEventListener/FlowEventPublisher on the Slot so
callers can subscribe via ExecuteOption.eventListener and receive
incremental events while a chain is running. ReActAgentComponent wires
agentscope reasoning/tool_result/summary/result events into this channel
when a listener is registered, keeping the non-listener path as a plain
blocking call.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Reflect the session management refactor from single sessionId to
(conversationId, agentKey) dual-key: update all hook signatures to
parameterless style with ctx(), add curl/wget to shell whitelist,
and remove obsolete SlotAttachmentTest.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
