How do I install the Lit Screen skill?

Run npx skillmds@latest add kennethkhoocy/lit-screen in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Lit Screen safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Lit Screen?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Lit Screen free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Lit Screen?

kennethkhoocy (@kennethkhoocy) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Lit Screen by kennethkhoocy — Agent Skill

Name: Lit Screen
Author: kennethkhoocy

Lit Screen

Stage 6 of the lit review pipeline: screen paper abstracts against the research prompt. The orchestrator's agent-driven flow runs this re-ranker on Opus subagents; a standalone run uses the in-script Claude Sonnet API fallback. Rates relevance 1-10, tags each paper as theoretical/empirical, identifies methodology, and classifies relationship to user's work. Only use this skill when explicitly requested -- e.g., the user says "run lit-screen", "lit-screen", or "/lit-screen". Do NOT auto-trigger on general literature review requests.

kennethkhoocy 7d18784 2 files · 36.8 KB Updated Sep 17, 2026

File contents

lit-screen (Stage 6 -- Abstract Screening)

Screen every paper's abstract against the original research prompt. In the orchestrator's agent-driven flow the relevance judgment is produced by Opus subagents through the --emit-tasks / --ingest-results seam (no API key); a standalone run uses the in-script Claude Sonnet API path instead. Produces a relevance score (1-10), rationale, and structured tags for each paper.

Usage

python ~/.claude/skills/lit-screen/scripts/lit_screen.py \
  --input stage5_merged.json \
  --query "your research prompt here" \
  -o stage6_screened.json

CLI Flags

Flag	Default	Description
`--input`	(required)	Input JSON from Stage 5 (dedup output)
`--query`	(required)	Research query/prompt to screen against
`-o, --output`	`stage6_screened.json`	Output JSON path
`--model`	`claude-sonnet-5`	Anthropic model ID (autonomous fallback)
`--concurrency`	`5`	Max simultaneous API requests (autonomous fallback)
`--emit-tasks PATH`	—	Agent-driven: write per-paper screening tasks and stop (no API)
`--ingest-results PATH`	—	Agent-driven: merge Opus screening results and write all outputs (no API)

Output Schema

Each paper gets these fields added:

{
  "screening_score": 8,
  "screening_rationale": "Directly examines board composition changes...",
  "paper_type": "empirical",
  "identification_strategy": "DiD",
  "relationship": "direct competitor"
}

Output Files

JSON: stage6_screened.json -- full paper list with screening fields
JSON: stage6_filtered.json -- papers with score >= 4 only
XLSX: stage6_screened.xlsx -- all papers, sorted by screening_score descending
XLSX: stage6_filtered.xlsx -- filtered papers (score >= 4), sorted by score descending
RIS: stage6_screened.ris -- for import into reference managers
BIB: stage6_screened.bib -- BibTeX entries for papers with score >= 5

Environment Variables

Var	Required	Description
`ANTHROPIC_API_KEY`	Fallback	Standalone-run screening (Sonnet API). The agent-driven flow screens with Opus subagents and needs no key.

Field Values

screening_score: 1 (irrelevant) to 10 (highly relevant); 0 = no abstract
paper_type: theoretical or empirical
identification_strategy: natural experiment, IV, DiD, RDD, structural, descriptive, N/A
relationship: foundational/must-cite, same method different context, same context different method, direct competitor, methodological reference, tangential

---name: lit-screendescription: Stage 6 of the lit review pipeline: screen paper abstracts against the research prompt. The orchestrator's agent-driven flow runs this re-ranker on Opus subagents; a standalone run uses the in-script Claude Sonnet API fallback. Rates relevance 1-10, tags each paper as theoretical/empirical, identifies methodology, and classifies relationship to user's work. Only use this skill when explicitly requested -- e.g., the user says "run lit-screen", "lit-screen", or "/lit-screen". Do NOT auto-trigger on general literature review requests.---# lit-screen (Stage 6 -- Abstract Screening)Screen every paper's abstract against the original research prompt. In theorchestrator's agent-driven flow the relevance judgment is produced by **Opussubagents** through the `--emit-tasks` / `--ingest-results` seam (no API key); astandalone run uses the in-script Claude Sonnet API path instead. Produces arelevance score (1-10), rationale, and structured tags for each paper.## Usage```bashpython ~/.claude/skills/lit-screen/scripts/lit_screen.py \ --input stage5_merged.json \ --query "your research prompt here" \ -o stage6_screened.json```## CLI Flags| Flag | Default | Description ||------|---------|-------------|| `--input` | (required) | Input JSON from Stage 5 (dedup output) || `--query` | (required) | Research query/prompt to screen against || `-o, --output` | `stage6_screened.json` | Output JSON path || `--model` | `claude-sonnet-5` | Anthropic model ID (autonomous fallback) || `--concurrency` | `5` | Max simultaneous API requests (autonomous fallback) || `--emit-tasks PATH` | — | Agent-driven: write per-paper screening tasks and stop (no API) || `--ingest-results PATH` | — | Agent-driven: merge Opus screening results and write all outputs (no API) |## Output SchemaEach paper gets these fields added:```json{ "screening_score": 8, "screening_rationale": "Directly examines board composition changes...", "paper_type": "empirical", "identification_strategy": "DiD", "relationship": "direct competitor"}```## Output Files1. **JSON**: `stage6_screened.json` -- full paper list with screening fields2. **JSON**: `stage6_filtered.json` -- papers with score >= 4 only3. **XLSX**: `stage6_screened.xlsx` -- all papers, sorted by screening_score descending4. **XLSX**: `stage6_filtered.xlsx` -- filtered papers (score >= 4), sorted by score descending5. **RIS**: `stage6_screened.ris` -- for import into reference managers6. **BIB**: `stage6_screened.bib` -- BibTeX entries for papers with score >= 5## Environment Variables| Var | Required | Description ||-----|----------|-------------|| `ANTHROPIC_API_KEY` | Fallback | Standalone-run screening (Sonnet API). The agent-driven flow screens with Opus subagents and needs no key. |## Field Values- **screening_score**: 1 (irrelevant) to 10 (highly relevant); 0 = no abstract- **paper_type**: `theoretical` or `empirical`- **identification_strategy**: `natural experiment`, `IV`, `DiD`, `RDD`, `structural`, `descriptive`, `N/A`- **relationship**: `foundational/must-cite`, `same method different context`, `same context different method`, `direct competitor`, `methodological reference`, `tangential`

Frequently asked questions