stage-tabs-builder

Generate booktabs LaTeX tables from fingerprinted evidence files with per-row source comments.

2|Updated Aug 3, 2026
One-click install
npx skills add https://github.com/wanghao9610/STAGE --skill stage-tabs-builder-wanghao9610
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: stage-tabs-builder
Source: https://github.com/wanghao9610/STAGE/tree/main/.qwen/skills/stage-tabs-builder
Command: npx skills add https://github.com/wanghao9610/STAGE --skill stage-tabs-builder-wanghao9610

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Academic results tables often contain numbers typed from memory or chat that cannot be traced back to any experiment, and this Skill eliminates that failure mode by compiling every table cell exclusively from fingerprinted evidence files under mates/. ## Core Features & Use Cases - Evidence-only table generation: Builds or updates one booktabs table under manus/tabs/ where every data row carries a % src: mates/<...>#<anchor> comment linking each number to its fingerprinted source. - Missing-data handling: Converts any number without registered evidence into a \todo{...} cell and opens an unsourced claim row in notes/claims.md so gaps stay visible to auditors and linters. - Staleness and verification gates: Runs import.sh --diff to detect stale evidence before typesetting and re-reads every cited number at its anchor before the table ships. - Use Case: After importing experimental results from a paired STAR research repo, ask for the main results table and receive a compile-ready LaTeX table where a reviewer can trace every metric back to its evidence fingerprint. ## Quick Start Run /stage-tabs-builder main_results to generate the main results table from the registered evidence in mates/.

Frequently Asked Questions about stage-tabs-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a LaTeX results table from experimental data?

Invoke the skill with a table ID, file slug, or purpose phrase, and it reads the fingerprinted evidence files under mates/, maps each row and column to an anchored number, and emits a booktabs table under manus/tabs/ with per-row source comments.

How do I keep table numbers traceable to their source experiments?

Each data row gets a % src: mates/<file>#<anchor> comment placed directly above it, so an auditor can walk from any cell to the exact evidence file and anchor. Numbers are re-read at their anchors before the table ships.

What happens when a table value has no evidence file?

The cell becomes a \todo{...} naming the missing measurement, and an unsourced claim row is opened in notes/claims.md. Hand-typed numbers from chat or memory are never written into cells; the user is routed to register the evidence first.

Can I use numbers from published papers as baseline rows?

Yes, but only after registering them through mates/manual/ with a MANIFEST entry via the evidence curation step. Even well-known literature numbers must carry a fingerprint before they can appear in a table cell.

Why does the table build stop when mates/ is empty?

An empty mates/ directory means there is no fingerprinted evidence to compile from, so the skill halts and routes to evidence import instead of starting a table on promises. This enforces the rule that evidence precedes typesetting.

Does the skill detect stale or outdated experimental evidence?

Yes, it runs import.sh --diff to compare source stamps of cited entries against upstream before typesetting. Stale evidence is reported and re-imported with user approval before its numbers ship.