New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Evidence Chain Builder

by OpenMinis in OpenMinis/MinisSkills

Score a claim against the evidence behind it. An agent skill from OpenMinis/MinisSkills.

MITAuto-check passedResearch & Science

Install Evidence Chain Builder

skills CLI
$ npx skills add OpenMinis/MinisSkills --skill evidence-chain-builder -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install OpenMinis/MinisSkills evidence-chain-builder --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/OpenMinis/MinisSkills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/evidence-chain-builder .claude/skills/evidence-chain-builder && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
evidence-chain-builder
GitHub stars
444
Token cost
~2.4k tokens
SKILL.md length
1,240 words
Files
4 (incl. scripts, references)
Skills in repo
49
Repo updated
First seen
Licence
MIT

At a glance

Score a claim against the evidence behind it. An agent skill from OpenMinis/MinisSkills.

  • Works in 6 steps: Separate the claim from the evidence… → Tag every item, and tag honestly. An… → Run the tool (command below). → …
  • The user wants to know whether something is actually supported — is that true
  • SKILL.md covers Overview, The one boundary that matters, Workflow and Command, plus 4 more sections
  • Runs Python scripts from its folder; calls python3

What it does

Evidence Chain Builder is an agent skill from OpenMinis/MinisSkills. Score a claim against the evidence behind it. Use this skill whenever the user wants to know whether something is actually supported — "is that true", "what's the source", "that sounds made up", "can you back that up" — or before repeating any number, statistic, market size, study result or benchmark that arrived without a citation, especially from a model or a summary. It separates the claim from its evidence, grades each item by source type, and reports an evidence score plus an explicit list of what is not…

Its SKILL.md is about 2.4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 6 other files, including scripts and reference files (for example `evals/evals.json`, `references/evidence-types.md` and `scripts/evidence_chain.py`). Compatibility notes: Python 3.8+ on PATH. Standard library only — nothing to install, no network, works offline. Writes nothing to disk unless the caller redirects the output.

It sits in Research & Science, covering Statistics and Citation management. The repository describes itself as: Skills collection for Minis. The licence is MIT.

When your agent uses it

  • The user wants to know whether something is actually supported — is that true
  • Whats the source
  • That sounds made up
  • Can you back that up —

Example prompts

  • “is that true”
  • “s the source”
  • “that sounds made up”
  • “/evidence-chain-builder”

Requirements

  • Python 3
  • Compatibility (from SKILL.md): Python 3.8+ on PATH. Standard library only — nothing to install, no network, works offline. Writes nothing to disk unless the caller redirects the output.

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

  1. Separate the claim from the evidence before running anything. Write the claim
  2. Tag every item, and tag honestly. An item is official only if it names a
  3. Run the tool (command below).
  4. Read the verdict band, not just the number. The bands are defined below.
  5. Name the gap concretely. Do not report "the evidence is weak". Report which
  6. Report the score and the gap together. A score without the unverifiable list

What it can do on your machine

Read from SKILL.md and the folder at commit ae8c5db. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (Python), which the agent can run.

    Shell commands in SKILL.md call:

    • python3

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

  • Compatibility

    Python 3.8+ on PATH. Standard library only — nothing to install, no network, works offline. Writes nothing to disk unless the caller redirects the output.

    From compatibility in the SKILL.md frontmatter.

Context cost

Evidence Chain Builder loads about 2.4k tokens when it runs, and up to ~4.1k if it reads all its reference files. Until then it costs about 216 tokens; SKILL.md has 1,240 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~216
When it runs · the whole SKILL.md, loaded when a task matches
~2.4k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~4.1k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from OpenMinis/MinisSkills at commit ae8c5db, republished under its MIT licence (© OpenMinis). 1,240 words, ~2,391 tokens.

Download SKILL.mdSave it as .claude/skills/evidence-chain-builder/SKILL.md (or your agent's skills folder). This skill also uses 3 other files; get the full folder from GitHub.
name
evidence-chain-builder
description
Score a claim against the evidence behind it. Use this skill whenever the user wants to know whether something is actually supported — "is that true", "what's the source", "that sounds made up", "can you back that up" — or before repeating any number, statistic, market size, study result or benchmark that arrived without a citation, especially from a model or a summary. It separates the claim from its evidence, grades each item by source type, and reports an evidence score plus an explicit list of what is *not* backed by independently verifiable sources. Also trigger when the user is preparing something that will be published, filed or relied on — a report, a filing, compliance material, a post quoting figures — and the claims need sources pinned to them first. Trigger on 证据链, 论断验证, 防幻觉, 信源可信, 有出处吗, 这句话有依据吗, 논거 검증, 근거, 裏付け, 出典.
compatibility
Python 3.8+ on PATH. Standard library only — nothing to install, no network, works offline. Writes nothing to disk unless the caller redirects the output.

Evidence Chain Builder

Overview

scripts/evidence_chain.py splits a claim from its evidence and scores how well the evidence carries it. Every evidence item is tagged with a source type, and the type decides whether that item counts:

typecounts as independently verifiable
official — a primary document, filing, standard, or the issuing body itselfyes
paper — peer-reviewed literature, or a dataset with a stated methodyes
data — raw measurements, logs, exports the user can re-runyes
internal — the user's own notes, records, or an undocumented internal numberno
assertion — the claim restated, someone's opinion, or no source statedno

The score is the share of items that count, as a percentage. internal and assertion are not "worth less" by taste — they are items nobody outside the conversation can check, and that is the property the score is measuring.

Full taxonomy, edge cases and tagging rules: references/evidence-types.md.

The one boundary that matters

This tool does not decide whether a claim is true. It grades the evidence, and nothing else.

That boundary is the whole point. A grader that also pronounced on truth would reproduce exactly the failure it exists to catch — a confident verdict unsupported by anything. So the tool reports what can be checked and what cannot, and stops.

Two consequences to carry into your answer:

  • A score of 100 does not mean the claim is correct. It means every item offered is the kind of thing that could be checked. Nobody has checked it yet.
  • A low score does not mean the claim is false. It means the claim is not established, which is a different and more useful statement.

Never upgrade "not established" into "false", and never upgrade a high score into "true". Both are the same mistake in opposite directions.

Workflow

  1. Separate the claim from the evidence before running anything. Write the claim as one sentence. If the user gave you three conclusions, that is three runs — a single chain will average unrelated things and hide the weak one.
  2. Tag every item, and tag honestly. An item is official only if it names a document or issuing body that exists independently of the conversation. "Our records show" is internal. "Studies suggest" with no study is assertion. If you cannot tell what an item is, it is assertion until the user says otherwise — guessing a stronger type defeats the tool.
  3. Run the tool (command below).
  4. Read the verdict band, not just the number. The bands are defined below.
  5. Name the gap concretely. Do not report "the evidence is weak". Report which items are unverifiable and what specifically would replace each one — "the failure-rate figure needs the QA report it came from, not a description of it".
  6. Report the score and the gap together. A score without the unverifiable list invites the reader to treat the number as a quality grade, which it is not.

Command

bash
python3 scripts/evidence_chain.py \
  --claim "Device X failure rate is below 0.5%" \
  --evidences "2025 QA audit report@official:QA dept" \
              "no complaints on record@internal:support" \
              "peer-reviewed cohort study@paper:JAMA"

Evidence strings take an optional @type:source suffix; one argument per item, or several joined with ||. No suffix means assertion.

flagmeaning
--claim TEXTthe claim to test
--evidences E1 E2 ...evidence items, optionally @type:source
--from FILE.json{claim, evidences:[{text, source, type}]} for longer chains
--threshold Nscore needed to count as supported (default 50)
--subject TEXTwhat the claim is about, printed in the header
--jsonmachine-readable output, for gating or downstream use
--fail-below-thresholdexit 1 when not supported; default is exit 0 either way

Exit codes: 0 the chain was scored, whatever the verdict; 2 usage or input error. Non-zero-to-signal-unsupported is opt-in, because a caller that treats any non-zero exit as "the command broke" would report a finding as an error.

Run python3 scripts/evidence_chain.py --help for the authoritative flag list.

Reading the result

conditionwhat you may say
no evidence suppliednothing is established. There is no chain to stand on.
score below thresholdDo not treat this claim as a conclusion. Say which items are unverifiable.
score at or above thresholdSupported to the stated threshold. The sources still need a human check.

The verdict text is deliberately phrased to survive being quoted out of context. Keep that phrasing when you relay it; do not round it up to "verified".

Threshold defaults to 50 because a chain where the majority of the weight is checkable is at least auditable. Raise it with --threshold when the claim will be published, filed, or acted on — 80 for anything going to a regulator or a customer. Say which threshold you used, since the verdict depends on it.

Show full SKILL.md (494 more words)Show less

What this cannot do

asked forstatus
Confirm a cited source existsNot checked. The tool never fetches anything.
Detect a fabricated citationNot possible here — a plausible-looking but invented reference scores as paper
Judge whether the claim is trueOut of scope by design; see the boundary above
Weigh study quality — sample size, bias, designNot modelled. paper means "a paper was named", not "a good paper"
Rank sources against each otherTypes are unordered beyond verifiable / not verifiable
Resolve conflicting evidenceTwo sources that disagree both score; the conflict is yours to surface
Handle a claim with several partsRun it per part; averaging hides the weak link

State these limits plainly when they bite rather than padding the answer. If the user needs source existence checked, that is a search task — do it as one, and treat the result as a new evidence item, not as the same chain.

Examples

A figure that arrived with no source. The user says "an analyst told me this market grows 30% a year". Run the chain with the figure as the claim and "analyst estimate@assertion" as its only evidence. The score is 0 and the verdict is that it is not established. Say that, and say what would fix it: the analyst's published note, or the underlying data. Do not soften it into "could be higher or lower".

Mixed chain, honestly reported.

bash
python3 scripts/evidence_chain.py --claim "Device X failure rate is below 0.5%" \
  --evidences "2025 QA audit report@official:QA dept" \
              "no complaints on record@internal:support" \
              "peer-reviewed cohort study@paper:JAMA"
# -> 67/100, 2 of 3 verifiable, threshold 50

Relay it as: two of the three items can be checked, the support log cannot, and the claim clears a 50 threshold but not an 80 one.

Gating a publish step.

bash
python3 scripts/evidence_chain.py --from claims.json --threshold 80 --fail-below-threshold

Exit 1 means at least one claim is not carried by verifiable evidence. Report which ones by name; do not just forward the exit code.

A non-English request. "这段有依据吗" needs no translation step — run the tool on the claim as written and answer in the user's language. The tool's own output is English; your reply is not.

Working on Minis

The sandbox makes this skill stronger than a plain script port, so use it:

  • Run it in the on-device shell. python3 scripts/evidence_chain.py ... works as-is in the Alpine sandbox; there is nothing to install.
  • Pin the artifacts you can reach. When an evidence item is a file the device can see — a contract, an export, a photo of a label, a record — add its sha256sum to the evidence text. A named document can be swapped later; a named document plus its digest cannot. That is the difference between "there is a report" and "there is this report".
  • Keep it on the device. The tool never uploads and never fetches, which is exactly what you want when the evidence is confidential. Do not compensate by pasting the evidence into a web search — that would undo the property that made this safe to run here.
  • Save the report into the workspace, not to a temp path, so the chain is still there in a later session when someone asks where the number came from.

© OpenMinis, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 3 other files (scripts, references) in evidence-chain-builder of OpenMinis/MinisSkills.

  • SKILL.md
  • evals/evals.json
  • references/evidence-types.md
  • scripts/evidence_chain.py

Open the folder on GitHubat commit ae8c5db

Compare with similar skills

Evidence Chain Builder next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Evidence Chain Builder compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Evidence Chain Builder this skillOpenMinis/MinisSkills444—~2.4kAutomated safety check: PassMIT
Reader First Technical Editlawve-ai/awesome-legal-skills842—~4.1kAutomated safety check: PassApache-2.0
Peer ReviewK-Dense-AI/claude-scientific-writer2.4k2 repos~3.1kAutomated safety check: NotesMIT
Manuscript Writing Reviewlabarba/sciwrite852—~2.4kAutomated safety check: PassCC-BY-4.0
Research Paper WritingRedWoodOG/Hermes-Desktop1776 repos~16kAutomated safety check: NotesMIT
Oleafly Review ManuscriptOleafly/Oleafly209—~2.3kAutomated safety check: PassMIT

Similar skills

  • Reader First Technical Edit

    lawve-ai/awesome-legal-skills

    Edit mathematically, empirically, or technically dense papers for readers outside the author’s specialty, including empirical legal scholarship, law-and-economics models, and legal-technology…

    842 GitHub stars~4.1k tokensUpdated 7 days ago
    Research & ScienceAuto-check passed

  • Peer Review

    K-Dense-AI/claude-scientific-writer

    Prepare evidence-bounded, constructive peer-review drafts and structured manuscript assessments.

    2.4k GitHub starsUsed in 2 repos~3.1k tokens
    Research & ScienceAuto-check: notes

  • Manuscript Writing Review

    labarba/sciwrite

    A skill your agent uses when asked to review, edit, or improve the writing quality of a scientific or engineering manuscript.

    852 GitHub stars~2.4k tokensUpdated 2 mo ago
    Research & ScienceAuto-check passed

  • Research Paper Writing

    RedWoodOG/Hermes-Desktop

    End-to-end pipeline for writing ML/AI research papers — from experiment design through analysis, drafting, revision, and submission.

    177 GitHub starsUsed in 6 repos~16k tokens
    Research & ScienceAuto-check: notes

  • Oleafly Review Manuscript

    Oleafly/Oleafly

    Run a referee-grade review of the manuscript in the open project without changing a single manuscript file.

    209 GitHub stars~2.3k tokensUpdated today
    Research & ScienceAuto-check passed

  • Literature Statistics

    aipoch/medical-research-skills

    Generate statistics for publication-year and journal distributions from local references or PDFs; use when you need standardized Year/Journal tables and a summary without any network access.

    2k GitHub stars~1.9k tokensUpdated 22 days ago
    Documents & OfficeAuto-check passed

More from OpenMinis/MinisSkills

All 49 skills in this repo

  • Android UI Automation

    OpenMinis/MinisSkills

    Automate Android apps that have no public API or web version by driving the UI layer through the Accessibility Service.

    444 GitHub stars~1.6k tokensUpdated 2 days ago
    Auto-check passed

  • Codex On Ish

    OpenMinis/MinisSkills

    Install and run OpenAI Codex CLI inside the Minis/iSH Alpine sandbox on iOS, where rustls TLS and async sockets are broken.

    444 GitHub stars~1k tokensUpdated 2 days ago
    Auto-check passed

  • Hyperframes CLI

    OpenMinis/MinisSkills

    HyperFrames CLI and Minis rendering. An agent skill from OpenMinis/MinisSkills.

    444 GitHub stars~3.5k tokensUpdated 2 days ago
    Auto-check passed

  • Whenpeak

    OpenMinis/MinisSkills

    Predict when a person's brain works best from their sleep, using the WhenPeak performance-intelligence API, and turn it into concrete scheduling advice.

    444 GitHub stars~2.3k tokensUpdated 2 days ago
    Auto-check passed

  • Openstreetmap Marker

    OpenMinis/MinisSkills

    Generate a mobile-first, immersive, Amap(Gaode)-style custom landmark marker HTML map for travel itinerary planning, place showcasing, location sharing, etc.

    444 GitHub stars~1.9k tokensUpdated 2 days ago
    Auto-check passed

  • PDF Converter

    OpenMinis/MinisSkills

    Convert documents to PDF via a typst-based pipeline. An agent skill from OpenMinis/MinisSkills.

    444 GitHub stars~2.1k tokensUpdated 2 days ago
    Auto-check passed

Categories

Questions about Evidence Chain Builder

What does Evidence Chain Builder do?

Score a claim against the evidence behind it. An agent skill from OpenMinis/MinisSkills. Evidence Chain Builder is an agent skill from OpenMinis/MinisSkills. Score a claim against the evidence behind it.

When should I use Evidence Chain Builder?

Evidence Chain Builder fits situations like: the user wants to know whether something is actually supported — is that true; whats the source; that sounds made up; can you back that up —.

How do I install Evidence Chain Builder in Claude Code?

Run `npx skills add OpenMinis/MinisSkills --skill evidence-chain-builder -a claude-code`. Or copy the skill folder (evidence-chain-builder in OpenMinis/MinisSkills) into .claude/skills/evidence-chain-builder in your project. Claude Code loads it when a task matches its description.

How do I install Evidence Chain Builder in Codex?

Run `npx skills add OpenMinis/MinisSkills --skill evidence-chain-builder -a codex`. Or copy the skill folder (evidence-chain-builder in OpenMinis/MinisSkills) into .agents/skills/evidence-chain-builder in your project. Codex loads it when a task matches its description.

Can I use Evidence Chain Builder in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add OpenMinis/MinisSkills --skill evidence-chain-builder -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/evidence-chain-builder, .gemini/skills/evidence-chain-builder, .github/skills/evidence-chain-builder and .opencode/skills/evidence-chain-builder in your project.

What does Evidence Chain Builder need to run?

Going by SKILL.md and its folder, Evidence Chain Builder needs Python for the scripts in its folder and the command-line tools its instructions call (python3). Our summary lists: Python 3. Compatibility (from SKILL.md): Python 3.8+ on PATH. Standard library only — nothing to install, no network, works offline. Writes nothing to disk unless the caller redirects the output. .

Does Evidence Chain Builder access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Evidence Chain Builder safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Evidence Chain Builder use?

Evidence Chain Builder is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Evidence Chain Builder use?

About 2.4k tokens (SKILL.md is roughly 9.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 1.7k tokens, read only when the agent opens those files.

What are the alternatives to Evidence Chain Builder?

Skills that share tags, products or a category with Evidence Chain Builder: Reader First Technical Edit (lawve-ai/awesome-legal-skills, 842 stars), Peer Review (K-Dense-AI/claude-scientific-writer, 2.4k stars), Manuscript Writing Review (labarba/sciwrite, 852 stars) and Research Paper Writing (RedWoodOG/Hermes-Desktop, 177 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Evidence Chain Builder?

OpenMinis (a GitHub organization) maintains it in OpenMinis/MinisSkills, which has 444 GitHub stars. The repository holds 49 skills in this directory. The repository was last updated on October 7, 2026.

Source: OpenMinis/MinisSkills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.