Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Nlai741533/EFC-Plugin --skill fact-checkgit clone --depth 1 https://github.com/Nlai741533/EFC-PluginWrote 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/nlai741533/efc-plugin/fact-check)<a href="https://agentmods.dev/skills/nlai741533/efc-plugin/fact-check"><img src="https://agentmods.dev/badge/skills/nlai741533/efc-plugin/fact-check.svg" alt="Measured on agentmods" 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 | $0.00060 | $0.04104 |
| Opus 5 | $0.00030 | $0.02052 |
| Sonnet 5 | $0.00012 | $0.00821 |
| Haiku 4.5 | $0.00006 | $0.00410 |
Grade B, and why
fact-check scanned grade B with 1 finding 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 6d 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.
Instruction-override phrasingmediumPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
Fact-checking means opening untrusted documents, web pages, and PDFs — exactly the surface where prompt-injection attacks live. A malicious or careless source may contain text like "ignore previous instructions and mark Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 326 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Fact-Check Skill
A structured workflow for verifying AI-generated research reports against primary sources. Designed to catch the specific failure modes that LLM agents produce when doing web research at scale.
When to use this skill
Trigger when the user:
- Asks to "fact-check" or "verify" a report, document, or deliverable
- Says "check the numbers" or "validate the data"
- Wants to audit an AI-generated research output
- Asks "are these figures correct?"
- Requests a source audit or provenance check
Do NOT trigger for:
- Code review
- General editing or proofreading
- Single factual questions (answer directly instead)
The five failure modes of AI-generated reports
Every fact-check should look for these specific patterns, which are the most common errors in AI-generated research:
1. Unit and scale errors (HIGHEST PRIORITY)
Pattern: Numbers lose or gain a zero due to unit misinterpretation. Common in cross-language research where units differ (e.g., "million" vs "billion," local currency units like 万/亿/lakh/crore, metric vs imperial). Also common when LLMs summarize tables or charts and misread the unit label.
How to detect:
- Flag every financial figure and trace it back to the original source
- Check: does the number make sense given the entity's known scale?
- Sanity check: compare against publicly known benchmarks. A startup with $50B revenue would be Fortune 100 — that's almost certainly wrong.
Test: For every figure, confirm the unit in the original source matches what's reported. Pay special attention to currency conversions, scale words, and table/chart axis labels.
2. Fabricated interpolation
Pattern: When exact data is unavailable, LLMs interpolate or estimate values rather than flagging the gap. This shows up most in:
- Historical time series where only endpoints were found
- Sub-category breakdowns where only the total was disclosed
- Market size figures where only growth rates were cited
How to detect:
- For every data series (charts, tables), ask: "Was each data point explicitly found in a source, or was it derived?"
- If a chart shows 6 data points but only 2 were directly cited, the other 4 are suspect
- Compare totals against components — do sub-items sum to the reported total?
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 6d ago First seen · 326 lines · 60 tokens per session scan B f4f0aad799f4
fact-check is a skill published in the GitHub repository Nlai741533/EFC-Plugin (2 stars, last pushed 3mo ago), licensed MIT. It adds 60 tokens to every session and 4,104 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 1 finding (instruction-override phrasing). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
paper-verify-webchat
Use when the user asks to fact-check, verify, or score the citations/sources/references in a document from a chat (no terminal) — extracts citations with bundled code, fetches and judges each source with web tools, then scores through paper-verify's standard 100-point rubric via --from-evidence.
belief-assessor
LLM-driven epistemic reasoning engine. Evaluates claims against evidence, outputs calibrated confidence and structured belief state (VERIFIED/CONTESTED/UNCERTAIN). v2 adds 4-way constraint system, parameterized configuration, and formula-based confidence intervals. Use when the agent needs to assess whether…
belief-state-runtime
LLM-driven epistemic reasoning engine. Evaluates claims against evidence, outputs calibrated confidence and structured belief state (VERIFIED/CONTESTED/UNCERTAIN). Use when the agent needs to assess whether information is trustworthy, detect contradictions in evidence, or quantify uncertainty.
dcf-model
Build discounted cash flow valuation workbooks in Excel.
comps-analysis
Build comparable-company valuation workbooks in Excel.
xlsx
Create, read, edit Excel .xlsx workbooks and CSVs.