How does the Table Filler prevent hallucinations?

The skill is governed by a strict guardrail that only allows it to pull data present in your evidence packs. It forbids the invention of facts and requires a citation for every single row.

Can I use this for long-form narrative inside table cells?

No, this skill specifically prevents long prose in cells. It is optimized for atomic facts and comparisons; narrative content should be moved to draft sections instead.

What happens if my evidence pack is missing data for the table?

The skill will identify this as an evidence gap. You should then enrich your evidence_drafts.jsonl file or adjust your schema rather than having the tool guess the information.

What files are required for this skill to function?

You need four primary inputs: a table schema (table_schema.md), subsection briefs, evidence drafts in JSONL format, and a BibTeX reference file (ref.bib).

Evidence-First Table Filler

Name: Evidence-First Table Filler
Author: WILLOSCAR

byWILLOSCAR

•

Learning & Documentation

Populates structured tables with citation-backed facts directly from evidence packs to eliminate placeholders and prose.

About

The Evidence-First Table Filler is a specialized Claude Code skill designed for high-integrity technical writing and research documentation. By mapping table schema definitions to verified evidence drafts, it generates clean, professional Markdown tables that are strictly limited to atomic facts and mandatory citations. This skill solves the common issue of AI-generated 'hollow' writing by enforcing a methodology where every row must be backed by a source, effectively preventing hallucinations and the use of placeholder text like 'TODO' or ellipsis in your final documentation.

Key Features

Zero-placeholder policy to ensure production-ready output
89 GitHub stars
Automated table generation from schema and evidence packs
Anti-prose guardrails to keep table cells concise and readable
Integration with research-units-pipeline evidence drafts
Strict citation enforcement using BibTeX keys for every row

Use Cases

Populating data-driven technical documentation from verified research evidence
Generating comparative technical feature tables in research papers
Creating literature review summaries with direct source citations

About

Key Features

Zero-placeholder policy to ensure production-ready output
89 GitHub stars
Automated table generation from schema and evidence packs
Anti-prose guardrails to keep table cells concise and readable
Integration with research-units-pipeline evidence drafts
Strict citation enforcement using BibTeX keys for every row

Use Cases

Populating data-driven technical documentation from verified research evidence
Generating comparative technical feature tables in research papers
Creating literature review summaries with direct source citations