evidence-ledger

To build an evidence ledger from a research paper, this Skill identifies observable artifacts and extracts span-anchored, hashed claims from text. It generates a manifest classifying artifacts by observability levels (L0, L1, L2) for accurate, reproducible analysis.

Automated paper analysis and artifact extraction support LaTeX, PDF, and text-only source files. The Skill processes these formats to construct a deterministic ledger by parsing the document content and identifying observable artifacts within the provided text.

Generating a deterministic artifact manifest requires Bash, Read, Write, Grep, and Glob tools, alongside specific Codex MCP access. This environment setup is necessary for the Skill to parse files and optionally enrich the ledger with semantic claims using an LLM.

Semantic enrichment for extracted paper works by optionally adding semantic claims using an LLM during an enrichment pass. This process supplements the deterministic, span-anchored hashed claims extracted directly from the text with additional context for audit preparation.

You can use this paper audit tool for LaTeX and PDF sources simultaneously, as it operates on LaTeX, PDF, and text-only sources. It constructs a deterministic evidence ledger across these formats to support comprehensive pre-audit preparation.