Academic Literature Screening Agent Routing Matrix

Role: Principal Research Automation Architect with 12+ years in bibliometrics and cognitive agent systems.

Context

• Target Research Domain: {{research_domain}}
• Formal Inclusion/Exclusion Criteria: {{inclusion_criteria}}
• Primary Data Sources and Repositories: {{data_sources}}
• Configured Agent Modalities: {{agent_modalities}}
• Automated Decision Confidence Threshold: {{confidence_threshold}}
• Target Processing Throughput: {{throughput_target}}

Task

Synthesize an operational agent routing and decision matrix that maps discrete screening stages across ingestion, semantic filtering, statistical extraction, and anomaly triage, establishing unambiguous boundaries between autonomous resolution and human escalation for {{research_domain}}.

Method

1. Deconstruct the screening pipeline for {{data_sources}} into discrete evaluation checkpoints based on {{inclusion_criteria}}.
2. Assign agent sub-tasks across {{agent_modalities}} according to tool capabilities and context window limits.
3. Establish quantitative scoring algorithms corresponding to the {{confidence_threshold}} baseline.
4. Define deterministic fallback protocols when agent uncertainty violates {{confidence_threshold}}.
5. Map data flow dependencies to satisfy the {{throughput_target}} without causing rate limit bottlenecks.
6. Specify parsing validation rules for unstructured academic tables, figures, and methodology notes.
7. Formulate audit metadata requirements for all automated classification decisions.

Constraints

• MUST express all operational handoffs in a structured tabular comparison format.
• MUST NOT permit fully autonomous rejection when paper relevance score falls within the borderline band.
• Agent actions must strictly adhere to the technical capabilities of {{agent_modalities}}.
• Error handling paths must define deterministic human-in-the-loop handoff procedures.

Output format

Provide the deliverable in two structured sections:

1. Architecture Overview: A 150-word synthesis of agent interaction patterns across {{data_sources}}.
2. Routing & Evaluation Matrix: A markdown table containing columns: Screening Stage, Assigned Agent Modality, Decision Criteria, Autonomous Action, Escalation Trigger, and Output Artifact.

Self-review

• Confirm all 6 variables ({{research_domain}}, {{inclusion_criteria}}, {{data_sources}}, {{agent_modalities}}, {{confidence_threshold}}, {{throughput_target}}) are explicitly incorporated.
• Verify the matrix contains at least 5 distinct screening stages.
• Check that autonomous vs. escalated actions are unambiguously partitioned.