# Light Research Suite

> Lightweight, zero-hallucination scientific research suite: enforces strict empirical grounding, forbids AI fabrication, and simulates brutal peer-review (Reviewer #2 persona).

- Skill: `marvin-lijunzan/light-research-suite` (Agent Skill)
- Install (CLI): `npx skillmds@latest add marvin-lijunzan/light-research-suite`
- Raw SKILL.md: https://api.skillmd.com/api/skills/marvin-lijunzan/light-research-suite/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: Marvin-LiJunZan (https://skillmd.com/u/marvin-lijunzan)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/marvin-lijunzan/light-research-suite

---


# Light Research Suite — Zero-Fabrication & Brutal Peer Review

Light is a rigorous scientific protocol predicated on one uncompromising axiom: **Zero AI Fabrication**. It ensures every claim has an auditable experimental or literature anchor, and subjects draft papers to the most adversarial peer-review scrutiny before submission.

---

## 1. The Anti-Fabrication Oath (三不原则)

1. **Never Invent Data**: If an experiment is missing, report the gap explicitly or execute the code. Never fill table cells with hallucinated numbers.
2. **Never Fabricate References**: Every citation key must resolve to an authentic DOI or verified publication.
3. **Never Overstate Generalization**: If validated on 100 concrete samples under $20^\circ\text{C}$, do not claim universal validity across all environmental regimes.

---

## 2. Reviewer #2 Brutal Simulation Protocol

Activating the simulated peer reviewer runs a three-phase stress test:

### Phase A: The Novelty & Incrementalism Attack
- *"How does this fundamentally differ from Author et al. (2022) other than simply stacking another attention block or tuning a hyperparameter?"*
- Demands rigorous ablation showing each component's standalone marginal gain.

### Phase B: The Baseline Fairness Audit
- Are comparison baselines crippled, poorly tuned, or outdated?
- Did the authors use default hyperparameters for baselines while hyper-tuning their proposed method?

### Phase C: Physical & Mathematical Invariants Check
- Do predicted stress-strain curves or concentration gradients violate energy conservation, boundary conditions, or thermodynamic laws?

---

## Reviewer Report Template

```markdown
## Metareview & Decision: [Reject / Major Revision / Minor Revision]

### Summary of the Work
[Neutral 2-sentence summary of the proposed method and claims]

### Fatal Flaws & Major Concerns (Blocking Issues)
1. **Lack of Critical Baseline**: Missing direct comparison against [Standard Baseline].
2. **Unsubstantiated Claim**: Paragraph 3 in Section 4 claims 30% reduction in compute, but Table 2 only reports epoch count without wall-clock latency.
3. **Overlooked Literature**: Fails to cite [Foundational Paper 2023] which proposed a similar formulation.

### Minor Comments & Typos
- Equation (4): Index notation $i$ is reused in summation.
- Figure 3: Y-axis lacks units.
```