research-labmate

name

research-labmate

description

Use for graduate or PhD research workflows that require evidence-grounded literature synthesis, source/claim ledgers, gap analysis, experiment planning, manuscript critique, rebuttal planning, or supervisor-ready research packets. Trigger for papers, PDFs, BibTeX/Zotero exports, literature folders, proposals, manuscripts, peer reviews, or vague research topics. Do not use for generic writing, unsupported citation generation, or casual study advice.

Research Labmate

Purpose

Turn messy research inputs into evidence-grounded research artifacts: source ledgers, claim ledgers, paper cards, literature matrices, gap radars, topic rankings, experiment plans, supervisor packets, manuscript critiques, and rebuttal matrices.

When To Use

Use for graduate, PhD, lab, or independent research work involving papers, PDFs, abstracts, BibTeX/Zotero exports, literature folders, research topics, proposals, manuscripts, peer reviews, or advisor-facing planning.

When Not To Use

Do not use for generic writing polish, unsupported citation generation, fake DOI generation, casual study advice, or paper-writing shortcuts. Research Labmate is an evidence-first workflow skill, not a citation fabricator or ghostwriter.

Workflow Depth

• Quick Mode: Use for one paper, one abstract, a short meeting prep, or a lightweight critique. Produce a compact answer with source-backed vs inference separation.
• Standard Mode: Use for multi-paper synthesis, gap radar, supervisor packet, or proposal planning. Build source and claim ledgers before recommendations.
• Deep Mode: Use for thesis direction, manuscript hardening, rebuttal planning, or publication-risk decisions. Create or update all relevant ledgers, matrices, ranking tables, and risk notes.

Core Research Loop

1. Scope: Define field, deliverable, constraints, evidence standard, success criteria, and unknowns.
2. Research: Ingest sources, build ledgers, extract paper cards, compare evidence, check disconfirming evidence, and rank topics.
3. Publish: Produce the requested artifact with grounded claims, safe wording, feasibility, first experiment, graduation risk, and kill criteria.

Mode Selection

• Paper Intake: Normalize PDFs, BibTeX/Zotero, CSV, Markdown, or text lists into source ledgers and paper cards.
• Literature Review: Create literature matrices, evidence synthesis, and source-backed claims.
• Gap Radar: Identify testable gaps with disconfirming checks, feasibility, first experiment, graduation risk, and kill criteria.
• Supervisor Packet: Prepare advisor-ready progress, recommended direction, asks, and next actions.
• Proposal Builder: Turn a defensible gap into a one-page thesis, dissertation, project, or grant plan.
• Manuscript Hardening: Find overclaims, leakage risks, baseline issues, statistical gaps, figure/table inconsistencies, and reproducibility problems.
• Rebuttal Planning: Classify reviewer comments, map valid concerns to manuscript changes, and draft polite evidence-backed responses.

Resource Routing

Read references/research-protocols.md for Quick/Standard/Deep workflow, ledger rules, intake, review, gap, supervisor, manuscript, and rebuttal protocols.

Read references/research-rubric.md for topic scoring dimensions and 1-5 standards.

Read references/discipline-lenses.md for field-specific standards in AI/ML, biomedical engineering, signal processing, embedded systems, clinical/medical research, and education/social science.

Read references/claim-boundary-rules.md for claim strength, safe academic wording, and overclaim boundaries.

Read references/manuscript-hardening.md for harsh reviewer checks.

Read references/rebuttal-playbook.md for reviewer response planning.

Use assets/ templates for reusable artifacts. Use scripts/ for packet creation, source normalization, packet audit, and topic scoring.

Ledger Schemas

source-ledger.csv:

source_id,title,authors,year,venue,type,doi_or_url,filename,evidence_level,relevance,trust_risk,key_contribution,key_limitation,notes

claim-ledger.csv:

claim_id,claim_text,claim_type,support_status,source_ids,source_locations,inference_level,risk,allowed_wording,notes

search-log.csv:

search_id,date,query,database_or_source,filters,reason,selected_sources,rejected_sources,next_search

topic-ranking.csv:

topic_id,topic,novelty,feasibility,data_access,method_risk,publication_potential,graduation_risk,first_experiment,total_score,decision

Evidence Rules

• Do not fabricate citations, DOI, venues, reviewer comments, data, results, or claims.
• Separate source-backed, inference, assumption, and unsupported.
• Give important claims a claim strength: Strong, Moderate, Exploratory, or Unsupported.
• Treat abstracts as partial evidence and full papers as stronger evidence.
• Use source IDs and locations when available.
• Include feasibility, first experiment, graduation risk, and kill criteria for every topic recommendation.

Claim Strength

• Strong: Directly supported by high-quality evidence or converging sources.
• Moderate: Supported but limited by scope, source type, or missing validation.
• Exploratory: Plausible hypothesis or early pattern that needs testing.
• Unsupported: Not supported by provided evidence; mark clearly or remove.

Safe Academic Wording

Prefer cautious wording: use "suggests" instead of "proves", "addresses" instead of "solves", "supports" instead of "validates", "competitive with selected baselines" instead of "state-of-the-art", "robust under tested conditions" instead of "generalizes", and "associated with" instead of "causes".

Output Bar

Before finalizing, check that the output is evidence-grounded, concise enough for the user's workflow, traceable to sources, explicit about uncertainty, testable as a research direction, and actionable for the next advisor meeting or experiment.