Borrowing it
Nothing to install: this file belongs to DanceNitra/agora. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/DanceNitra/agora/main/.claude/skills/verify-claims/SKILL.mdgit clone --depth 1 https://github.com/DanceNitra/agoraWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/dancenitra/agora/verify-claims)<a href="https://agentmods.dev/skills/dancenitra/agora/verify-claims"><img src="https://agentmods.dev/badge/skills/dancenitra/agora/verify-claims/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/dancenitra/agora/verify-claims"><img src="https://agentmods.dev/badge/skills/dancenitra/agora/verify-claims.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00163 | $0.01922 |
| Opus 5.5 | $0.00065 | $0.00769 |
| Sonnet 5.5 | $0.00033 | $0.00384 |
| Haiku 4.5 | $0.00016 | $0.00192 |
Grade A, and why
verify-claims scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verify Claims — adversarial pre-publish fact-check
What this does
Takes a draft (a post .md, a Crucible note, an outreach reply, pasted text) and, before it goes public,
independently verifies every checkable claim against its PRIMARY source — not a blog summary, the
actual paper/dataset/filing. It returns truthful verdicts (CONFIRMED / PARTIALLY CONFIRMED / UNVERIFIED /
FALSE), the exact corrections, a verification banner, and a claim-safety guide, then patches the draft.
This is the operational form of Agora's hard rule "verify measured numbers vs the source before any
public citation" (we once posted an unreproduced "+144%"; never again). It is the same adversarial pass
that, run inside storm-research, caught a misattributed arXiv ID, a 1,692→1,689 sample-size error, and a
fragile vendor stat — all before publish.
Self-contained: built-in Agent (general-purpose), WebSearch/WebFetch, Read/Edit. No external services.
When to use (hard gate)
Run this and pass it BEFORE: rendering/publishing any post, recording a public Crucible entry, posting a gated outreach reply or press piece, or approving any outward action that cites a number or source. A piece that has not passed this is not ready to publish.
Process
1. Extract the checkable claims
Read the target. Pull EVERY independently-checkable assertion:
- each cited number / statistic / effect size / percentage / dollar figure / date / sample size,
- each named study / paper / author / venue / arXiv-or-DOI id,
- each direct quote or "X said/found Y",
- each claim of priority ("first to…", "nobody has…").
List them. Our OWN measured numbers count too — re-check them against
server/.lab.json/ the lab script / re-run, not against a vault note (a number in a note is NOT verified).
1b. Run the CONSTRUCTION gate on every artifact behind one of OUR numbers (mandatory, blocking)
python -X utf8 tools/publish_gate.py <draft.md> # audits every runnable model the draft links
python -X utf8 tools/publish_gate.py <probe.py> ... # or name the artifacts directly
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2mo ago First seen · 116 lines · 163 tokens per session scan A 45454d2338f6
verify-claims is a skill published in the GitHub repository DanceNitra/agora (3 stars, last pushed 7d ago), licensed MIT. It adds 163 tokens to every session and 1,922 once invoked, about $0.0007 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
verify-claims compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| verify-claimsDanceNitra/agora | Skill | 3 | 7d ago | MIT | A | 163 | Claude Code |
| Evaluationagentic-in/elephant-agent | Skill | 584 | 2mo ago | none | A | 25 | Claude Code, Codex |
| create-custom-graderNVIDIA/SkillEvaluator | Skill | 643 | yesterday | Apache-2.0 | A | 35 | Claude Code, Codex |
| api-callerNVIDIA/SkillEvaluator | Skill | 643 | yesterday | Apache-2.0 | A | 29 | Claude Code, Codex |
| calculatorNVIDIA/SkillEvaluator | Skill | 643 | yesterday | Apache-2.0 | A | 32 | Claude Code, Codex |
| text-analyzerNVIDIA/SkillEvaluator | Skill | 643 | yesterday | Apache-2.0 | A | 42 | Claude Code, Codex |
| company-braincaura-ai/caura | Skill | 549 | yesterday | Apache-2.0 | A | 108 | Claude Code |