Academic Research

Academic Research

This is the Codex entry router for the vendored research suite under ars/.

Defaults

• Domain default: geography, human geography, GIScience, and health geography.
• Chinese default: Simplified Chinese. Use Traditional Chinese only when the user explicitly asks for it.
• NORA is not an installed skill here. Use ../shared/field-context/ only as a local reference pack when domain grounding is needed.
• Read ../shared/research-calibration.md for analysis planning, experiment execution, question evaluation, or manuscript-readiness decisions. When the user is developing an analysis or manuscript and gives no stage, default to first-draft + proportionate rigor + incremental-allowed contribution + balanced compute.
• Do not import submission, formal peer-review, systematic-review, or reviewer-response standards into first-draft analysis unless a named validity risk makes them immediately necessary.
• For reading, outlining, drafting, or revision, read references/selective-quality-control.md. Check load-bearing claims first, accept bounded PDF/manuscript reading scopes, use claim skeletons before uncertain prose, and prefer local patches over whole-section rewrites.
• For question framing, analysis planning, paper architecture, Results, Discussion, or revision, also read ../shared/advantage-led-research-narrative.md. Build author-side work around the leading advantage, winning arena, and decisive evidence; surface only material constraints that change the claim or route.

Use the Research Continuity record in ../shared/advantage-led-research-narrative.md: preserve question, advantage state, fair comparison, claim/experiment IDs, evidence version, and retellable sentence across stages. Update only fields affected by the current task; reuse project records rather than starting another ledger.

Execution scope

Read execution and approval scope when using internal workflows. Phase completion returns control to the main agent; continue the authorized task while preserving applicable explicit approvals.

First rule

Do not load the whole suite by default. Start from one workflow entry file and then load only the needed agent, reference, or template files.

Routing boundaries

• Use good-question when the main decision is whether a question is important, tractable, falsifiable, or worth pursuing.
• Use good-story when results or manuscript materials already exist and the main task is to organize them into an evidence-supported narrative.
• Use academic-advisor for proposal diagnosis, pre-submission assessment, and journal targeting.
• Use peer-review for a formal external-reviewer report.
• Keep the internal question-refinement and reviewer workflows available only when the user explicitly invokes this skill or needs an end-to-end pipeline.

Workflow router

• Deep research, literature review, systematic review, meta-analysis, research question refinement: ars/deep-research/WORKFLOW.md
• Academic paper writing, outline, abstract, revision, citation formatting, disclosure, formatting guidance: ars/academic-paper/WORKFLOW.md
• Peer review, reviewer simulation, editorial decision, re-review: ars/academic-paper-reviewer/WORKFLOW.md
• End-to-end research-to-paper pipeline: ars/academic-pipeline/WORKFLOW.md
• Experiment planning, statistical interpretation, reproducibility, human study protocol: ars/experiment-agent/WORKFLOW.md

Mode routing

Prefer /academic-research mode=<mode> or the equivalent natural-language request. Supported modes include:

• plan
• outline
• abstract
• lit-review
• citation-check
• disclosure
• format-convert
• revision-coach
• revision
• full

Legacy ars-* tokens may be interpreted as these internal modes when supplied after /academic-research; they are not separate slash-callable skills.

If the user only has a broad topic or tentative title but no clear research question, use good-question first unless the user explicitly requests the end-to-end academic-research pipeline.

Codex mapping

• Treat upstream Claude/agent dispatch wording as phase prompts to execute inline.
• Do not install or run Claude hooks.
• Treat WORKFLOW.md files as the actual internal entrypoints.
• Use ../shared/field-context/geography-publication-readiness.md for geography-specific contribution, method-fit, theory, argument, and title checks.
• Apply ../shared/research-calibration.md before interpreting contribution magnitude, deciding robustness scope, or scheduling expensive analysis.
• Use references/selective-quality-control.md for impact-first review, bounded novelty language, evidence-led Introduction recalibration, and advisory claim-drift checks.
• Use ../shared/advantage-led-research-narrative.md to prevent reviewer-first drafting, experiment accumulation without an argument job, and project-diary Results structures.
• Use ../shared/field-context/academic-writing.md, spatial-methods.md, geoai-domain.md, environmental-health.md, and journal-templates/ only when relevant.
• For projects that use dynamic greenspace exposure and Space-Time Constraints, read references/space-time-greenspace-terminology.md before drafting, revising, or auditing terminology. Treat it as a project/domain calibration reference, not a universal replacement for field-standard wording.