What Outputs Are Generated During the RESEARCH Stage of the ARS Pipeline: 7 Critical Artifacts Explained

The RESEARCH stage of the ARS (Academic Research Skills) pipeline generates seven distinct artifacts—including the RQ Brief, Methodology Blueprint, and INSIGHT Collection—that serve as the foundation for all downstream academic writing.

The ARS pipeline is an open-source academic research automation framework maintained at Imbad0202/academic-research-skills. During Stage 1 – RESEARCH, the system orchestrates multiple specialized agents to transform raw inputs into structured research assets. These outputs are defined in docs/ARCHITECTURE.md and tracked programmatically by the state_tracker_agent before being handed off to the academic-paper skill for writing.

Core Artifacts Produced During RESEARCH

The RESEARCH stage implements a deep-research skill that emits exactly seven deliverables. Each artifact is produced by a dedicated agent and follows strict validation protocols before advancing to Stage 2.

RQ Brief (Research Question Brief)

The research_question_agent generates the RQ Brief, a concise document containing a FINER-scored research question, scope boundaries, and 2–3 sub-questions. This artifact establishes the intellectual foundation for the entire pipeline.

The brief is stored as a markdown file and referenced in the pipeline status template at academic-pipeline/templates/pipeline_status_template.md under the placeholder {rq_brief_path}.

Methodology Blueprint

The research_architect_agent produces the Methodology Blueprint, a comprehensive YAML or markdown file detailing the research paradigm, specific methods, data strategy, analytic framework, and validity criteria. This blueprint functions as the technical specification for the study design.

According to the source code in docs/ARCHITECTURE.md, this artifact appears in the Stage × Dimension matrix as a required output before the pipeline can proceed.

Annotated Bibliography (S2-Verified)

The bibliography_agent and source_verification_agent collaborate to create the Annotated Bibliography, which includes systematic literature search results with evidence-hierarchy grading and predatory-journal checking. The "S2-verified" designation indicates that these sources have been pre-validated for quality and relevance.

When a Material Passport contains a literature_corpus[], the bibliography integrates with the Search-Strategy Report to document evidence selection.

Synthesis Report

The synthesis_agent assembles the Synthesis Report, containing thematic synthesis, gap analysis, and a knowledge-integration narrative. This document translates raw literature into actionable research insights and identifies voids in the current academic landscape.

INSIGHT Collection

The devils_advocate_agent and risk_of_bias_agent harvest the INSIGHT Collection—a JSON or markdown file containing high-impact findings, open questions, and risk-of-bias flags. These items serve as cognitive guardrails during the writing phase, ensuring critical evaluation persists throughout the pipeline.

Search-Strategy Report with PRE-SCREENED Block

When processing a Material Passport that includes a literature_corpus[], the search_strategy_agent generates a report containing a PRE-SCREENED block. This section explicitly records which corpus entries were kept, skipped, or filtered out during the selection process, ensuring full auditability of the evidence base.

Decision-Heavy Checkpoint Prompt

Unlike other artifacts, the Decision-Heavy Checkpoint is an interactive output that pauses the pipeline execution. Defined in docs/ARCHITECTURE.md section 2.3, this prompt requires user confirmation that the RQ Brief and Methodology Blueprint are acceptable before the system transitions to Stage 2 (Analysis).

How Outputs Are Tracked and Stored

The pipeline uses a status template system to monitor artifact generation. The orchestrator populates academic-pipeline/templates/pipeline_status_template.md with filesystem paths to each output:

# Stage 1 RESEARCH    [{status_icon}] {status_text}

#   - RQ Brief:          {rq_brief_path}

#   - Methodology:      {methodology_path}

#   - Bibliography:     {bib_path}

#   - Synthesis:        {synthesis_path}

#   - Insight collection:{insight_path}

When rendered, the template displays concrete paths:

Stage 1 RESEARCH    ✅ Completed
  - RQ Brief:          /tmp/pipeline/rq_brief.md
  - Methodology:      /tmp/pipeline/methodology.yaml
  - Bibliography:     /tmp/pipeline/annotated_bibliography.md
  - Synthesis:        /tmp/pipeline/synthesis_report.md
  - Insight collection:/tmp/pipeline/insight_collection.json

Downstream agents access these artifacts programmatically via scripts/state_tracker_agent.py, which maintains the state graph and ensures file handles are passed correctly to the academic-paper skill.

Key Source Files for RESEARCH Outputs

File	Role
docs/ARCHITECTURE.md	Defines the Stage × Dimension matrix enumerating all RESEARCH outputs
deep-research/SKILL.md	Orchestrates the agent flow and checkpoint logic
academic-pipeline/templates/pipeline_status_template.md	Human-readable status template displaying output paths
agents/research_question_agent.md	Generates the RQ Brief
agents/research_architect_agent.md	Generates the Methodology Blueprint
agents/bibliography_agent.md	Generates the Annotated Bibliography
agents/synthesis_agent.md	Generates the Synthesis Report
agents/devils_advocate_agent.md & agents/risk_of_bias_agent.md	Produce the INSIGHT Collection

Summary

The RESEARCH stage of the ARS pipeline produces seven rigorously defined outputs:

• RQ Brief: FINER-scoped research question with sub-questions
• Methodology Blueprint: Technical specification of research design
• Annotated Bibliography: S2-verified literature with quality grading
• Synthesis Report: Thematic analysis and gap identification
• INSIGHT Collection: Critical findings and bias flags
• Search-Strategy Report: PRE-SCREENED evidence audit trail
• Decision-Heavy Checkpoint: Human-in-the-loop validation gate

These artifacts are defined in docs/ARCHITECTURE.md, tracked by state_tracker_agent.py, and stored in the temporary pipeline directory before handoff to academic writing agents.

Frequently Asked Questions

What is the RQ Brief in the ARS pipeline?

The RQ Brief is a structured document created by the research_question_agent that contains a FINER-scored research question, clearly defined scope boundaries, and 2–3 specific sub-questions. It serves as the intellectual anchor for all subsequent pipeline stages and must be validated by the user at the Decision-Heavy Checkpoint before proceeding.

How does the Annotated Bibliography get verified?

The bibliography_agent performs the initial literature search, while the source_verification_agent conducts S2 verification—including evidence-hierarchy grading and predatory-journal detection. This dual-agent approach ensures that only high-quality, credible sources enter the synthesis phase.

What happens if the Decision-Heavy Checkpoint is rejected?

If the user rejects the RQ Brief or Methodology Blueprint at the Decision-Heavy Checkpoint, the pipeline loops back to the research_question_agent or research_architect_agent for iterative refinement. The system does not advance to Stage 2 until both artifacts receive explicit user confirmation, as implemented in the checkpoint logic defined in docs/ARCHITECTURE.md section 2.3.

Where are RESEARCH stage outputs physically stored?

Outputs are written to a temporary pipeline directory (typically /tmp/pipeline/ or a configurable path) with standardized filenames: rq_brief.md, methodology.yaml, annotated_bibliography.md, synthesis_report.md, and insight_collection.json. The state_tracker_agent.py maintains handles to these paths for downstream consumption by the academic-paper skill.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}