SKILL

Cs Litreview

/cs:litreview <research-question> — Academic literature orientation. Grill-me intake (question + framework + depth), Consensus recon, framework checkpoint, sequential budget-allocated searches (5/10/20), 8-section .docx research guide output. Sibling of /cs:pulse (research pack).

Published by @Alireza Rezvani·0 agent reads / 30d·0 saves·

/cs:litreview — Academic Literature Orientation

Command: /cs:litreview <research question>

The cs-litreview persona produces a strategically planned mini literature review as an 8-section .docx research guide.

When to Run

  • Starting research on an unfamiliar field
  • Writing a paper that needs grounding in current literature
  • Mapping the "lay of the land" before committing to a research direction
  • Want a curated reading list with key authors + foundational papers + gaps

When NOT to Run (use Consensus directly)

  • Looking for ONE specific paper (just search Consensus)
  • Quick lookup with no need for synthesis
  • Field you already know well and just need a recent papers list

Forcing Intake (3 Questions, One at a Time)

QAsksDefault if forcing-choice
Q1Research question (1-2 sentences, specific)refuses vague; "AI in medicine" gets pushed back once
Q2Framework: PICO / SPIDER / Decomposition / Hybrid / You-pick"you pick" (skill recommends from Q1)
Q3Tentative depth: Quick (5) / Standard (10) / Deep (20)re-confirmed at post-Phase-2 checkpoint

What You Get

After Phase 0 intake + Phase 1 recon + Phase 2 framework + interactive checkpoint + Phase 3 searches:

research_guide_<topic>_<date>.docx with 8 sections:

  1. Topic Overview — single tight paragraph
  2. Start Here — Priority Reading Order — 5-7 hyperlinked papers (best-review → foundational → frontier → gap)
  3. How the Field Got Here — chronological narrative + timeline table
  4. Sub-area Guides — one per sub-area (4 parts each: synthesis / key papers / search terms / boolean strings)
  5. Key Research Groups — top 3-5 authors/groups with representative papers
  6. Open Questions & Gaps — methodological / population / conceptual
  7. Bibliography — alphabetical, hyperlinked, every inline citation matches
  8. Audit Log — search table + counts + detected plan tier

Interactive Checkpoint (Mid-Run)

After Phase 2 (framework selected, sub-areas generated), the skill halts with a forcing-options prompt:

Framework breakdown:
| {Component} | How it maps to your topic | Proposed sub-area |
|---|---|---|
| Population | ... | Sub-area 1: ... |
| Intervention | ... | Sub-area 2: ... |
| Comparison | ... | Sub-area 3: ... |
| Outcome | ... | Sub-area 4: ... |
| Cross-cutting | ... | Sub-area 5: ... |

Confirm depth (plan-tier detected: free / ~10 results per search):
  1. Quick scan (5 searches)
  2. Standard review (10 searches)
  3. Deep dive (20 searches)

Sub-area options:
  - Looks good — proceed
  - Adjust: add sub-area on [X]
  - Adjust: replace [Y] with [Z]
  - Restart with different framework

This is the last cheap moment to correct course before search budget is consumed. Skill refuses to start Phase 3 without explicit user choice.

Discipline (Research-Pack Convention)

  • One intake question per turn. Never bundle.
  • Sequential Consensus calls. 1 q/sec rate limit. NEVER parallelize.
  • Plan-tier detected at first search, reported at checkpoint.
  • Halt at checkpoint. No Phase 3 without confirmation.
  • Source discipline — cite only THIS session's Consensus results. Training knowledge labeled [Not from Consensus].
  • Three-count tracking — searches / unique papers / cited.
  • Retry once after 3s — then log. 3 consecutive failures → stop.

Workflow

# Phase 0 intake (Q1-Q3 one at a time)
python ../skills/litreview/scripts/citation_tracker.py --action start --session NAME
python ../skills/litreview/scripts/framework_recommender.py --question "<Q1>"

# Phase 1 recon (1 Consensus search; record sent + received)
# Phase 2 framework + sub-area generation
# CHECKPOINT — wait for user

# Phase 3 searches (sequential, 1 q/sec, budget per tier):
#   5/10/20 searches across sub-areas + review + era-gated + follow-up

# Phase 4 cross-search aggregation + DOCX
python ../skills/litreview/scripts/cross_search_aggregator.py --session NAME
# Generate DOCX via Node.js docx library
python3 -c "import zipfile,sys; zipfile.ZipFile(sys.argv[1]).testzip()" output.docx  # zip-integrity check; then confirm required sections present

python ../skills/litreview/scripts/citation_tracker.py --action close --session NAME

Trigger Phrases (auto-invoke without /cs:)

  • "litreview on [topic]"
  • "literature review on [topic]"
  • "I'm starting a literature review on X"
  • "I'm writing a paper on X"
  • "help me research X"
  • "I'm doing research on X"
  • "can you help me research X"

Do NOT trigger for: single one-off paper searches — that's a plain Consensus search.

Anti-Patterns Rejected

  • Parallelizing Consensus calls
  • Skipping the interactive checkpoint
  • Padding thin results with training knowledge
  • Defaulting to non-PICO without justification
  • Citing papers in chat that didn't come from Consensus this session
  • Hardcoding plan tier instead of detecting
  • Skipping era-gated searches in standard/deep budgets
  • Skipping cross-search intelligence (repeat-hits, recurring authors)
  • Truncating Consensus URLs

Related

  • Agent: cs-litreview
  • Skill: litreview
  • Source spec: megaprompts/09-litreview-megaprompt.md
  • Sibling: /cs:pulse (research pack)
  • Future siblings: /cs:grants, /cs:patent, /cs:dossier, /cs:syllabus

Version: 1.0.0 Source: Path-B direct conversion of megaprompts/09-litreview-megaprompt.md

Bundled with this artifact

2 files

Reference files that ship alongside this artifact. Agents pull these in only when the task needs them.

More on the bench

SKILL▲0

Scvi Tools

Deep generative models for single-cell omics. Use when you need probabilistic batch correction (scVI), transfer learning, differential expression with uncertainty, or multi-modal integration (TOTALVI, MultiVI). Best for advanced modeling, batch effects, multimodal data. For standard analysis pipelines use scanpy.

K-Dense-AI
data-science-ml+2
⊙ 0
SKILL▲0

Scientific Visualization

Meta-skill for publication-ready figures. Use when creating journal submission figures requiring multi-panel layouts, significance annotations, error bars, colorblind-safe palettes, and specific journal formatting (Nature, Science, Cell). Orchestrates matplotlib/seaborn/plotly with publication styles. For quick exploration use seaborn or plotly directly.

K-Dense-AI
data-science-ml+2
⊙ 0
SKILL▲0

Experimental Design

Design experiments and studies BEFORE data is collected — choosing a design, randomizing, blocking, and laying out treatment combinations so the results will actually be interpretable. Use whenever someone is planning a study, asks how to assign subjects/samples to groups, mentions randomization, blocking, stratification, controls, factorial or fractional-factorial designs, design of experiments (DOE), screening many factors, response-surface optimization, crossover or repeated-measures or split-plot designs, cluster/group randomization, Latin squares, plate layouts, batch/run-order effects, replication vs. pseudoreplication, or sequential/adaptive/group-sequential designs. Trigger this even for informal phrasings like "how should I set up this experiment", "how do I avoid confounding", "what's the best way to test these 6 factors", or "assign these mice to conditions". For computing the sample size or power once the design is chosen, use statistical-power; for analyzing data already collected, use statistical-analysis.

K-Dense-AI
academic+2
⊙ 0