night_owl_research_agent

Adversarial iterative review loop with generator-evaluator separation. Up to 4 rounds of independent review, improvement, and re-evaluation. Persists state to output/REVIEW_STATE.json for recovery. Stop: score ≥ 7.5/10 on all dimensions, or 4 rounds. Writes full history to output/AUTO_REVIEW_REPORT.md.

Deploy and run experiments for ML/DL training (local, remote, or Modal GPU) AND spatial data science / GIScience experiments (local, data-driven). Reads from output/refine-logs/EXPERIMENT_PLAN.md and output/refine-logs/FINAL_PROPOSAL.md, writes to output/experiment/. Use when user says "run experiment", "deploy experiment", "execute experiment plan", or needs to launch training / spatial analysis jobs.

Turn a refined GIScience / remote sensing / spatial data science proposal into a detailed, claim-driven experiment roadmap. Use after `refine-research`, or when the user asks for a detailed experiment plan, ablation matrix, spatial evaluation protocol, run order, compute budget, or paper-ready validation that supports the core problem, novelty, simplicity, and contribution.

Reviews the manuscript produced by `paper-draft` (in `output/manuscript/`) as a demanding IJGIS / ISPRS JPRS reviewer-editor, cross-checks it against `output/PAPER_PLAN.md` and its evidence artifacts, then revises it into a stronger draft. Produces a reviewed manuscript, a revised manuscript, a structured review report, a prioritized issue log, a revision log, claim-risk notes, journal-fit notes, and next-loop priorities. Supports full, section-scoped, and mode-scoped review (structural / argument / novelty / methods / results-discussion / journal-fit / language / integrated). Safe on partial or skeletal drafts. Never fabricates results, citations, or figures.

Guideline-driven spatial analysis skill. Given a research question and data context, provides decision frameworks for selecting appropriate spatial methods, diagnostics, and interpretation strategies. Adapts to available data, spatial units, and analytical objectives — Claude Code determines the optimal workflow. Use when user says "spatial analysis", "analyze spatial data", "run spatial regression", "check for clustering", "map this", or needs to go from a research question to a complete spatial analysis.

Validates a manuscript against target journal requirements before submission. Checks word count, section structure, figure/table count, reference format, and geo-specific reporting standards. Writes a PASS/FAIL checklist report to output/reports/.

Complete 4-stage end-to-end research pipeline. Orchestrates idea-discovery-pipeline → deploy-experiment → auto-review-loop → generate-report. Reads RESEARCH_PLAN.md (or BRIEF.md as fallback) for context that overrides $ARGUMENTS.

Generate and rank research ideas given a broad direction. Use when user input "brainstorm ideas", "generate research ideas", "what can we work on", or wants to explore a research area for publishable directions.

The full pipeline for idea generation. It generates 8-12 novel research ideas from literature gaps and evaluates each on novelty, feasibility, and domain fit. Orchestrates lit-review → generate-idea → novelty-check → idea-review → experiment-design-pipeline to go from a broad research direction to a validated, pilot-tested idea with a refined proposal and experiment plan. Produces output/IDEA_REPORT.md plus refinement and experiment artifacts.

Discover, evaluate, and download publicly available datasets from the internet. Infers data needs from a research question or task, selects authoritative sources, downloads reproducibly, validates file integrity, and documents provenance. Pauses for user input when authentication, API keys, or major tradeoffs require a decision. Use when user says "download data", "get data", "find a dataset", "I need boundary files", "download census data", or needs any external dataset for analysis.

Consolidates literature review, idea discovery, refined proposal, experiment plan, experiment results, and automated review into a single NARRATIVE_REPORT.md that is rich enough to drive the downstream paper-writing-pipeline (paper-plan → paper-figure-generate → paper-draft → paper-review-loop → paper-convert).

Converts the final Markdown manuscript from `paper-draft` / `paper-review-loop` into a submission package for the target venue — modular LaTeX (one file per section), compiled PDF, and Word `.docx`. Venue is read from `output/PAPER_PLAN.md` (or argument) and routed through a small YAML profile. Does not rewrite prose, score, or invent citations.

Transforms output/PAPER_PLAN.md into a journal-quality Markdown manuscript draft for GIScience, GeoAI, spatial data science, and remote sensing venues (IJGIS, ISPRS JPRS, RSE, TGIS, AAG Annals). Consults referenced literature, experiment, figure, and claim artifacts; supports full drafts, partial drafts, and skeleton drafts depending on readiness. Never fabricates results, metrics, or citations — produces a claim-to-evidence map and coverage-gap report alongside the manuscript.

Generates publication-quality figures and diagrams from output/PAPER_PLAN.md for GIScience, GeoAI, and remote sensing journals (IJGIS, ISPRS JPRS, RSE, TGIS). Decides per-figure whether to produce reproducible code-generated plots/maps or structured prompts for external image-generation models (nano banana, ChatGPT image). Produces figure files, source scripts, captions, manifest, and prompt artifacts. Never fabricates results — uses only evidence from project files.

Monitors running spatial experiments. Checks output files, log files, and process status. Categorizes results as OK, STALLED, FAILED, or COMPLETE. Fires alerts by appending to output/PROJ_NOTES.md. Run every 15 minutes during Stage 3 of research-pipeline.

Produces a detailed, venue-aware paper outline from all available upstream artifacts (RESEARCH_PLAN, lit review, idea report, refined proposal, experiment plan/results, narrative report). Creates section-by-section plan with word budgets, claim-evidence matrix, figure/table plan, and citation scaffolding. Uses journal templates from templates/ for venue-specific formatting.

Full paper writing pipeline. Orchestrates paper-plan → paper-figure-generate → paper-draft → paper-review-loop → paper-covert to go from upstream research artifacts to a polished, submission-ready manuscript package (Markdown → LaTeX → PDF + DOCX). Use when user says "write paper pipeline", "paper writing", or wants the complete paper generation workflow.

Run an end-to-end workflow that chains the skills `refine-research` and `experiment-design`. Use when the user wants a one-shot pipeline from vague research direction to focused final proposal plus detailed experiment roadmap, or asks to build a pipeline, do it end-to-end, or generate both the method and experiment plan together.

Retrieves papers from local folder or ArXiv and Semantic Scholar using domain-aware keyword expansion, builds synthesis matrix, identifies gaps. Calls tools/arxiv_fetch.py and tools/semantic_scholar_fetch.py. Writes to output/paper-cache/ and output/LIT_REVIEW_REPORT.md.

Validates that a research idea is genuinely novel vs. existing literature. Searches ArXiv, Semantic Scholar, and WebSearch for near-duplicate work. Produces a novelty verdict and evidence. Run on each idea from idea-discovery-pipeline before investing in experiments.

Turn a vague research direction into a problem-anchored, elegant, frontier-aware, implementation-oriented method plan via iterative GPT-5.4 review. Use when the user says "refine my approach", "decompose this problem", "refine research plan", or wants a concrete research method that stays simple, focused, and top-venue ready instead of a vague or overbuilt idea.

Get a deep critical review of research idea from GPT via Codex MCP. Use when user says "review my idea", "help me review", "get external review", or wants critical feedback on research ideas, papers, or experimental results.