Matilde v1: verifiable-citations engine (axes 1-3) + Hermes plugin, SOUL & SKILL#1
Merged
juniperbevensee merged 2 commits intoJun 16, 2026
Conversation
…lugin Matilde's flagship: verify a citation along four axes — existence, metadata-match, retraction, URL-liveness — into a composite verifiability score + verdict (verified|warnings|retracted|not_found|unverifiable). - engine/citations.py: pure, injected-I/O verifier (stdlib only). Multi-source existence (Crossref -> OpenAlex -> DataCite) so real arXiv/Zenodo/DataCite DOIs are never mislabeled "fabricated". Retraction via Crossref Retraction Watch (updated-by/update-to/relation) + OpenAlex is_retracted. URL-liveness with a Wayback fallback. - hermes-plugin: matilde_verify_citation / _verify_bibliography / _check_retraction tools; no API key required. - 51 offline unit tests + 3 live API integration tests (MATILDE_LIVE=1), validated against real records (Wakefield retraction; arXiv DataCite DOI). - Honest framing: "verifiable", not "provably correct" — claim-support grounding (GROBID + SciFact/SemanticCite) is the v2 axis. - README states the academia/science use case explicitly (discoverability); sanitize.config tuned for study/manuscript particulars. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… checklist - docker/SOUL.Matilde.md: research-assistant identity centered on verify-don't- assume, calibrated confidence, never-fabricate; tiered source/action policy. - hermes-skill/SKILL.md: SCOPE→GATHER→VERIFY→SYNTHESIZE→REPORT lifecycle with the citation tools wired in; honest note that "verified" = metadata-checked, not claim-support-grounded (the v2 axis). - Remove SETUP.md (instantiation complete). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
juniperbevensee
deleted the
dev/juniperbevensee/matilde-citation-engine-v1
branch
June 16, 2026 02:21
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
The first real increment of Matilde — the academia/science customization of the Hermes use-case-package template — built from adversarially-verified research.
Flagship: the verifiable-citations engine
Verifies a citation along four axes → composite score + verdict (
verified | warnings | retracted | not_found | unverifiable):is_retractedCorrectness highlight: multi-source existence means real arXiv/Zenodo/DataCite DOIs are never mislabeled "fabricated" — a bug caught and fixed against the live APIs during development (Crossref alone 404s on arXiv
10.48550/*DOIs).Honest framing: verifiable, not provably correct. Claim-support grounding (does the cited passage back the claim?) is the probabilistic v2 axis (GROBID + SciFact/SemanticCite).
Agent surface (no API keys needed)
matilde_verify_citation·matilde_verify_bibliography·matilde_check_retractionTests
MATILDE_LIVE=1), validated against real records (Wakefield 1998 retraction; arXiv DataCite DOI).Also
SOUL.Matilde.md(verify-don't-assume identity) +SKILL.md(research methodology).Roadmap (researched, not yet built)
OpenNeuro/BIDS dataset analysis · meta-science stat-rechecking (statcheck/GRIM) · manuscript writing → LaTeX (Pandoc/Quarto) · v2 claim-support grounding.
🤖 Generated with Claude Code