MengkunLiang/DIG-ResearchOS

literature-scout

LLM guidance for designing literature search profiles and queries.

ideation

LLM guidance for ResearchOS T4 idea generation.

literature-synthesis

LLM guidance for staged literature synthesis.

manuscript-writing

LLM guidance for staged academic manuscript writing in ResearchOS.

research-reboost

Deterministically compile ResearchOS Pre-T5 research artifacts into a source-traceable, schema-valid external_executor/handoff_pack.json. This is the contract used by T5-REBOOST-GATE when it re-boosts project, literature, hypothesis, novelty, risk, and…

literature-evidence-scout

Search for and synthesize verified academic literature evidence for a scoped research question, preserving source identifiers and section-level provenance. Use before drafting, ideation, or citation repair when the user needs real source leads rather than…

t4-evolution

Start, resume, inspect, or safely pause the native T4 Evolutionary Idea workflow without editing its population, scoring, lineage, or selection files by hand.

baseline-reproduction

Reproduce, repair, and validate required ResearchOS baselines against the locked experiment protocol using Phase B approved resources and a root-owned iteration plan. Use when `research-execution` dispatches D1 because a required baseline has no valid…

code-and-protocol-review

Independently review ResearchOS external-executor implementation changes and experiment protocols before smoke, small-scale, or formal runs. Use when `research-execution` dispatches Phase D3 after code, config, adapter, dataset split, metric, evaluation,…

context-alignment

Validate and normalize the ResearchOS T5 external-executor handoff before resource preparation or experiments. Use when `research-execution` dispatches Phase A, when handoff/control inputs changed, when resuming with a stale alignment checkpoint, or when…

evidence-packaging

Compile one pinned final ResearchOS external-executor evidence snapshot into a realized method package, an evidence-bound final framework figure, a reproducible result figure/table inventory, and bidirectional method/code/config/result/visual mappings. Use…

experiment-design

Convert the confirmed ResearchOS external-executor scope and Phase B resource readiness into a versioned, claim-bound, budget-feasible experiment plan with a locked protocol fingerprint, mechanism-linked ablations, an execution DAG, predeclared interpretation…

experiment-run

Execute and checkpoint approved ResearchOS T5 external-executor runs with complete provenance. Use when `research-execution` dispatches a reviewed iteration run, or when an approved smoke, small-scale, formal, ablation, robustness, diagnostic, or efficiency…

implementation

Implement an approved ResearchOS method, baseline adapter, bounded repair, training/evaluation entrypoint, configuration, ablation switch, diagnostic switch, logging path, or test in a controlled versioned worktree. Use when `research-execution` dispatches…

method-refinement

Compile the approved ResearchOS method intent and current root-owned iteration decision into a versioned, implementation-ready method specification with stable module contracts, training/inference flows, config and ablation interfaces, failure modes,…

module-attribution

Attribute observed ResearchOS experiment behavior to implemented modules and bounded mechanisms using direct ablations, controlled diagnostics, implementation mappings, subset evidence, and explicitly labeled correlational hints. Use when `research-execution`…

research-execution

Orchestrate or resume the complete ResearchOS external-executor workflow from T5 handoff through resource readiness, claim-bound experiment planning, build-review-run iterations, diagnosis, attribution, evidence packaging, and Writer Handoff. Use when Codex…

resource-and-baseline-preparation

Prepare and statically validate the datasets, benchmarks, baseline implementations, evaluation code, preprocessing assets, checkpoints, environments, protocols, and references required by the confirmed ResearchOS external-executor scope. Use when…

result-diagnosis

Diagnose one ResearchOS experiment iteration from durable run records and raw metric evidence. Use when `research-execution` dispatches Phase E1 because a new usable experiment-run set has not yet been diagnosed, or when a prior diagnosis became stale after…

writer-handoff

Compile the final external-executor state into `executor_research_report.md` and validate the complete ResearchOS writing handoff. Use after evidence-packaging and all experiment skills finish, when final terminal `executor_status.json`, `result_pack.json`,…

paper-write

Create an evidence-bounded academic LaTeX manuscript from a research brief or an existing ResearchOS outline. Use for Chinese or English paper drafting when the user needs section-level prose, provenance-aware citations, mechanical manuscript audits, and a…

citation-graph-explorer

Explore one-hop references and related works from a DOI or OpenAlex work, then produce a bounded, provenance-aware citation neighborhood for literature expansion or baseline discovery. Use when a researcher wants to snowball from seed papers without treating…

citation-library-curator

Audit and curate a user-supplied BibTeX library into a separate, source-traceable reference candidate set with duplicate, malformed, unresolved, and verification-status reporting. Use before citation provenance audit, paper writing, or submission preparation.

claim-evidence-map

Map a batch of proposed manuscript or research claims to exact ResearchOS note sections, evidence levels, permitted wording, and evidence gaps. Use when a researcher needs a claim-by-claim source map before drafting, review, revision, or submission.

cross-domain-idea-studio

Develop evidence-bounded cross-domain research directions by retrieving bridge evidence, extracting transferable mechanisms, auditing migration risk, and submitting candidates to a human selection gate before hypothesis compilation. Use when a researcher…

domain-synthesis-studio

Build a complete, evidence-bounded domain synthesis from an initial research brief and available ResearchOS literature artifacts, optionally searching for missing literature before producing method families, mechanisms, tensions, boundaries, and survey or…

draft-evidence-repair

Audit a manuscript or section for unsupported claims, missing citations, stale evidence mappings, and overly strong wording, then prepare a traceable repair package without changing project facts. Use when a researcher needs to repair evidence before review,…

hypothesis-compiler

Compile a user-selected research direction into falsifiable hypotheses, mechanisms, boundary conditions, and an evidence-aware test plan. Use after idea selection and before any external experiment execution.

idea-fanout-jury

Generate, compare, and score evidence-calibrated research directions from a problem brief and ResearchOS literature artifacts. Use when a researcher wants a transparent, creativity-enabled idea fan-out and human-readable jury report before choosing a…

literature-comparison-studio

Turn a researcher’s comparison question plus DOI, title, PDF, paper-note, or source list into a traceable pairwise, corpus, or method-family comparison with evidence anchors and a decision-ready report. Use when paper-comparison alone is too manual because…

literature-evidence-matrix

Build a compact, source-traceable comparison matrix from a bounded set of ResearchOS paper notes, PDF note cards, or identifier records. Use when a researcher needs a review-ready evidence table before synthesis, survey writing, or idea selection.

literature-gap-map

Build an evidence-bounded map of unresolved research questions, boundary conditions, and missing comparisons from existing ResearchOS literature artifacts. Use before ideation or a related-work section when a researcher needs to distinguish retrieval gaps…

literature-review-studio

Run a guided, evidence-aware literature review workflow from review question and seed materials through query planning, source retrieval, citation expansion, reading coverage, synthesis, taxonomy readiness, and an optional Survey handoff. Use when a…

paper-comparison

Compare two or more note cards, DOI/arXiv/OpenAlex identifiers, direct PDF URLs, uploaded PDFs, or an explicitly scoped topic-plus-count request on a user-defined research question, method, evidence, limitation, and contribution axis.

paper-identifier-resolver

Resolve DOI, arXiv, OpenAlex, Semantic Scholar, or title identifiers into source-traceable paper records and a conservative BibTeX candidate library. Use when a researcher has pasted paper identifiers rather than PDFs and needs verified metadata,…

paper-outline

Build an evidence-bounded academic paper outline, contribution map, and readiness report from a research brief. Use before drafting when the user needs a clear argument structure without turning assumptions or unverified literature into paper claims.

paper-peer-review

Produce an evidence-aware, venue-sensitive peer-review report and prioritized revision plan for an academic LaTeX manuscript without rewriting the manuscript. Use after a draft exists and before revision, compilation, or submission.

paper-polish

Polish a Chinese or English academic LaTeX manuscript while preserving verified claims, citations, numbers, and author-controlled scope. Use after a draft exists when the user needs clearer prose, stronger structure, and a traceable change report rather than…

paper-reading-workbench

Organize one or more researcher-provided PDFs, DOI, arXiv identifiers, or existing paper records into prioritized reading cards, section-level answers, and a cross-paper learning summary. Use when a researcher wants a practical reading workspace rather than a…

paper-revision

Revise an academic LaTeX manuscript from reviewer comments with an evidence-aware response letter and traceable change log. Use when the user has a draft plus reviews and needs revisions that distinguish accepted edits, rebuttals, and requests requiring new…