Repositories / OpenClaw
Focused tour of agent execution, memory search, and tool orchestration; not a whole-project map.
Prepares workspace, prompt, runtime plan, authentication, and trajectory context for an execution attempt.
Stored evidence from this immutable analysis snapshot
const runtimePlan = buildAgentRuntimePlan({provider,modelId,model: runtime.effectiveModel,modelApi: runtime.effectiveModel.api,harnessId: runtime.agentHarness.id,harnessRuntime: runtime.agentHarness.id,preparedAuthPlan: runtime.activePreparedAuthPlan,
config: params.config, workspaceDir, agentDir, agentId: workspaceResolution.agentId, thinkingLevel: mapThinkingLevelForProvider(runtime.thinkLevel), extraParamsOverride: { ...params.streamParams, fastMode: attemptFastMode }, });