facts
Skill1 starsChanged 3 months agoMITView on GitHub
Audit the document against the facts file and record new facts in it. Use when the user wants to check that numbers, dates, and terminology are consistent across sections, or wants to add or update an authoritative figure. Triggers on "check the numbers", "are the figures consistent", "update the facts", "add this to facts", "fact check".
6 other projects use a version of this.
Adopted by
Most recently taken up by
- persiyanov/skillstheir own version, first seen Nov 15, 2026
- ai-search-guru/getcito-worlds-first-open-source-aio-aeo-or-geo-tooltheir own version, first seen Nov 8, 2026
- ansvisor/ansvisortheir own version, first seen Nov 1, 2026
- av/harbortheir own version, first seen Oct 8, 2026
- VladTheProgrammer/claude-asimovtheir own version, first seen Oct 8, 2026
- backnotprop/brotheir own version, first seen Jul 12, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- av/harbor★ 3.3ksame name
- ai-search-guru/getcito-worlds-first-open-source-aio-aeo-or-geo-tool★ 432same name
- ansvisor/ansvisor★ 241same name
- backnotprop/bro★ 182same name
- persiyanov/skills★ 1same name
- VladTheProgrammer/claude-asimov★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- figureSkill
- full-reviewSkill
- ingestSkill
- onboardSkill
- plan-sectionSkill
- researchSkill
- reviewSkill
- save-materialSkill
- statusSkill
- teachSkill
- verify-sourcesSkill
- writeSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
When given a new figure, date, definition, or decision:
- Very specific: line 18
Check whether docs/facts.md already covers it.
- Very specific: line 20
If the fact contradicts something already written in sections/, grep for the old value and list every location that now needs updating.
- Distinctive: line 28
Check each against the facts file, and report:
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/facts — Keep one version of the truth
docs/facts.md is the single source of truth for every number, date, sample size, definition, and terminology decision in the project. Sections quote it. When a figure changes, it changes here first and propagates outward.
This matters because the most damaging errors in a long document are not bad arguments, they are a sample size that is 4,312 in the methodology, "roughly 4,000" in the results, and 4,132 in the abstract. An examiner who finds one of those stops trusting all of them.
Mode A — Record a fact
When given a new figure, date, definition, or decision:
- Check whether
docs/facts.mdalready covers it. If it does and the value differs, do not silently overwrite: show both, ask which is correct, and ask what changed. - Record it with everything needed to defend it later: the value, where it came from (query, file, notebook, source document), when it was produced, and any caveat on how it should be reported.
- If the fact contradicts something already written in
sections/, grep for the old value and list every location that now needs updating. Do not edit the sections without being asked.
Mode B — Audit
Given a section, or the whole document:
- Read
docs/facts.md. - Extract every number, percentage, date, sample size, currency amount, and defined term from the target files.
- Check each against the facts file, and report:
- Contradictions — the document says one thing, the facts file another. Both locations, both values.
- Unrecorded — a figure appears in the document that the facts file does not carry. It needs a source, or it needs to go.
- Drift — the same quantity reported at different precision or rounding in different places, or a term used in a variant form.
- Stale — a fact in the file that no section uses any more, which usually means an analysis was superseded and the file was not updated.
- Report as a table with file and line for each. Do not fix anything unless asked.
Structure of the facts file
Group by what the facts are about, not by which section uses them, since most are used in several. A workable set of headings:
# Facts, decisions, and assumptions
Single source of truth. When a number appears in more than one place, this file is authoritative.
Last updated: <date>
## 1. Data and sample
<Sizes, date ranges, collection method, inclusion and exclusion rules, and what each snapshot or extract is called.>
## 2. Results
<Every reported figure, with its interval or error where it has one, and the analysis that produced it.>
## 3. Definitions and terminology
<The exact wording for each recurring term, its preferred form, and the near-synonyms that are not to be used.>
## 4. Decisions
<Methodological choices, when they were made, why, and what was rejected. This section answers most defence questions.>
## 5. Assumptions and known limits
<What the work takes for granted and where the evidence stops.>Where the project has several datasets, snapshots, or study waves that could be confused, give each one an unambiguous name and use that name everywhere. Conflating two of them is the single most common factual error in an empirical document, and the hardest to spot once it is written.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

