monarch-initiative/mechmaker/evidence-references

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Find, fetch, quote and validate the evidence behind a claim in DaTMech. Use when adding or fixing an `evidence` item, when a snippet fails reference validation, when choosing between sources, or when asked whether a claim is supported.

Install

npx vskill@latest install monarch-initiative/mechmaker/evidence-references

SKILL.md

Evidence and references

Every evidence item is checked by linkml-reference-validator. It fetches the cited source, caches it under references_cache/, and requires the snippet to appear in it verbatim.

Writing an item

  1. Find a primary source that states the claim. Prefer the paper that

showed it over a review that repeats it.

  1. Fetch it and read it:

``bash just fetch-reference PMID:<n> ``

  1. Copy the snippet from the fetched text. ... joins two exact spans.

[...] marks an editorial note. Nothing else may differ.

  1. Add it with just add-evidence. --at names where it goes: record for

the record's own evidence, or a path to one entry of a section, such as <section>[0] or <section>[2].<subsection>[1]:

``bash just add-evidence data/zip_areas/<stem>.yaml --at <place> \ --ref PMID:<n> --snippet "<verbatim>" --source <EvidenceSourceEnum value> \ --explanation "<how it bears on the claim>" --model <model-id> ``

That is a dry run: it fetches the source if needed, checks the quote with the same validator just validate uses, fills reference_title, and prints the diff. Add --apply to write it. It also appends a curation_history event, and refuses a quote it cannot find, a duplicate, or a place the schema gives no evidence.

The item it writes:

```yaml

  • reference: PMID:<n>

reference_title: <as fetched> supports: SUPPORT evidence_source: <EvidenceSourceEnum value> snippet: <verbatim> explanation: <how it bears on the claim> ```

An item written another way, by hand or by a conversion script, gets its title from the cache with just fill-titles <file> --apply.

  1. Validate the whole record once its evidence is in:

``bash just validate data/zip_areas/<stem>.yaml ``

supports

ValueWhen
SUPPORTThe quote states or directly shows the claim.
PARTIALIt supports part of the claim, or with a caveat that matters.
REFUTEIt contradicts the claim. Keep it. Open a CONTROVERSY discussion.
NO_EVIDENCERelevant context that does not bear on the claim.
WRONG_STATEMENTThe claim was found wrong. Kept for provenance.

When validation fails

  • Snippet not found. Your quote differs from the source. Fetch again and

copy. If the text is in the full paper but only the abstract is cached, quote the abstract or find an open-access copy. Do not paraphrase.

  • Reference not found. The identifier is wrong. Search PubMed for the

title. Never adjust digits until something resolves.

  • Unknown prefix. The reference type is not supported by the validator.

Use a PMID or DOI for the same work.

When no source exists

Do not reach. Write the claim as a KNOWLEDGE_GAP discussion with rationale saying what evidence would settle it. Sketch a proposed_experiments entry if one is obvious.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
5
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/monarch-initiative/mechmaker/evidence-references/badge)](https://verified-skill.com/skills/monarch-initiative/mechmaker/evidence-references)

Something wrong with this skill or its page? Tell us.