skill-facts
Skill0 starsChanged 29 days agoMITView on GitHub
Gives every claim in a skill a source and a checked date, and removes advice nobody can source. Use before publishing a skill that gives advice or numbers.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skill-benchSkill
- skill-checkSkill
- skill-competeSkill
- skill-createSkill
- skill-descriptionSkill
- skill-evalSkill
- skill-improveSkill
- skill-learnSkill
- skill-packageSkill
- skill-planSkill
- skill-pluginSkill
- skill-publishSkill
- skill-releaseSkill
- skill-repoSkill
- skillSkill
- skill-tidySkill
- skill-tokensSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Read the skill and its references.
- Distinctive: line 40
When a skill is about to be released, open every source again.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Facts with sources
A skill that tells the model "posts with links get less reach" or "the limit is 2,200 characters" is only as good as the day that was true. This finds each claim, sources it, dates it, and moves it into a registry the skill points at, so one refresh updates every skill that uses it.
1. Find the claims
Read the skill and its references. A claim is anything that could be checked and could change: a number, a limit, a format, a platform's behaviour, a "best practice". List each with where it sits.
The skill-fact-checker agent can do steps 1 and 2 in a fresh context and write the list.
2. Source each one
For each, find the best source and record how strong it is:
| Level | When |
| official-confirmed | The platform states it in its own documentation |
| official-guidance | The platform recommends it |
| strong-study | Two or more independent, large measurements agree |
| single-study | One measurement, named with its sample |
| anecdote | A secondary source, or practitioner reports |
Open each source and confirm it says what the claim says before you record it. A link you didn't read isn't a source. Record the address and today's date as checked.
3. The registry
Write the facts to facts/<topic>.json in the skill or plugin: one entry per claim with id, the claim in plain words, a machine-readable value when there is one, the evidenceLevel, and sources with url, title and checked. In the skill, say the claim and name the fact's id, so the next refresh knows where it's used.
4. What has no source
A claim you couldn't source either comes out of the skill or is reworded as what it is ("some practitioners report..."). Never raise a claim's level because it's useful. Report the list: sourced, reworded, removed.
Refreshing
When a skill is about to be released, open every source again. A source that changed means the claim, and every skill that names it, changes too.
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.