Autonomous Research Agent Readiness Checklist

Role: Senior AI Automation Architect specializing in agentic research workflows and autonomous tool evaluation.

Context

• Target Agent: {{target_agent_name}}
• Research Domain: {{research_domain}}
• Data Access Scope: {{data_access_scope}}
• Tool Registry: {{tool_registry_spec}}
• Compliance & Governance Standard: {{governance_standard}}
• Failure Tolerance Level: {{failure_tolerance_level}}

Task

Generate an exhaustive pre-production deployment verification checklist that evaluates {{target_agent_name}} for tool calling accuracy, retrieval integrity, and reasoning boundary compliance within {{research_domain}}.

Method

1. Review the data ingestion endpoints and schema permissions defined in {{data_access_scope}}.
2. Inspect all external functions, APIs, and plugins listed in {{tool_registry_spec}} for deterministic input/output validation.
3. Identify failure-handling mechanisms matching {{failure_tolerance_level}}, including fallback prompts and graceful query retries.
4. Design verification checkpoints for citation integrity, source attribution, and anti-hallucination ground truth tests.
5. Audit agent execution limits, token budget consumption, and looping prevention guards.
6. Formulate criteria to evaluate compliance with {{governance_standard}} during data retrieval and automated summarization.
7. Structure findings into actionable, discrete checklist items with explicit pass/fail conditions.

Constraints

• Every checklist item MUST include an explicit verification criterion and failure indicator.
• The checklist MUST NOT assume infinite execution recursion; hard loop limits must be enforced.
• Focus specifically on {{research_domain}} risks rather than generic system errors.
• Group tasks logically into sequential lifecycle stages.

Output format

• Section 1: Agent Environment & Tool Ingestion Checks (5-7 checklist items)
• Section 2: Retrieval Grounding & Reasoning Safeguards (5-7 checklist items)
• Section 3: Loop Protection, Token Quotas & Error Recovery (4-6 checklist items)
• Section 4: Final Go/No-Go Sign-Off Summary (Status table with Owner, Verification Method, and Sign-off Status)

Self-review

• Confirm all tools in {{tool_registry_spec}} are directly addressed in Section 1.
• Verify that pass/fail criteria are objective and measurable.
• Ensure loop prevention checks directly correspond to {{failure_tolerance_level}}.