Agent Claude Harness¶
Component id: agent_claude
Source contract: BlockContract for agent_claude.
Overview¶
Beta, review-pending Anthropic Claude agent with tool calling, autonomous tool-use loops, retry policy, circuit breaker, and lifecycle management. Block contract verification_method=tier1_review_pending; @component_maturity("beta").
When to use:
- Single-shot Claude inference with tool calling (
infer()) - Autonomous multi-step agent execution: call → observe → execute → repeat (
infer_loop()) - Production deployments requiring retry, circuit breaker, and structured logging
Good at¶
- single-shot Claude inference with tool calling
- autonomous multi-step tool-use loops (call → observe → execute → repeat)
- deployments needing retry, circuit breaker, and structured logging
Not suited for¶
- enforcing a central spend cap (paid calls are consent-gated, not quota-enforced)
- running where the Claude Code CLI is absent from PATH or unauthenticated
Mitigates¶
Known Failure Modes¶
review_pending: operator_review
State Surface¶
- Persistence:
run - Queryable via:
snapshot_export - Retention:
run
Extension Points¶
training_signal_inputs:('agent_routing_outcomes',)validation_gate:unit_plus_codex_reviewallowed_mutation_surfaces:('prompts', 'routing_policy')
Verification¶
- Method:
tier1_review_pending - Human review gate:
False - Fallback block:
none