reality-check
Skill4 starsChanged 5 months agoMITView on GitHub
Verification gate for status claims. Nothing gets called sent, replied, scheduled, shipped, merged, deployed, done, or closed until it has been checked against the live thing. Fires automatically before any status list, pipeline summary, weekly recap, standup, or follow-up plan reaches the user, and on demand when they say "/reality-check", "are you sure", "did I actually", "verify that", "check that before you tell me". Applies to outreach, deals, tickets, deploys, and anything a tracker, a note, or earlier in the session claims is done.
14 other projects use a version of this.
Adopted by
Most recently taken up by
- scoobydrew83/skillstheir own version, first seen Nov 3, 2026
- microsoft/amplifier-smart-tool-reality-checktheir own version, first seen Oct 17, 2026
- Valeriy1980/claude-skillstheir own version, first seen Oct 13, 2026
- SubliminalCoding/bilbroswagginz-skillstheir own version, first seen Sep 27, 2026
- devpilotX/skillstheir own version, first seen Sep 22, 2026
- creationisttest-git/startup_studiotheir own version, first seen Sep 12, 2026
Other versions14 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 14
- dazuck/operator-skills★ 17same name
- EmanueleMinotto/minottobot★ 4same name
- microsoft/amplifier-smart-tool-reality-check★ 2same name
- quatico-solutions/agent-skills★ 2same name
- dazuck/claude-code-skills★ 1same name
- evanmcgoogan/skin-in-the-game★ 1same name
- creationisttest-git/startup_studio★ 0same name
- dabodamjan/app-idea-to-spec★ 0same name
- devpilotX/skills★ 0same name
- ibangtam/reality-check-skill★ 0same name
4 more not listed
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 repository17 other files it gives its agents.
- ai-writing-qaSkill
- build-captureSkill
- build-modeSkill
- fitfoSkill
- flow-mapSkill
- generate-system-mapSkill
- imessageSkill
- leanSkill
- meta-promptSkill
- no-bullshitSkill
- plain-englishSkill
- pressure-testSkill
- problem-logSkill
- swipeSkill
- video-breakdownSkill
- viral-hook-creatorSkill
- working-with-claudeSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 12 edited versions we read keep these.
5
Distinctive
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 18
Before any of these words reach the user, open the actual artifact:
- Distinctive: line 50
If you did not check it, say you did not check it.
- Distinctive: line 51
Never say something is done because it should be done by now.
- Distinctive: line 52
Never infer that someone replied, or that something shipped, because time has passed.
- Distinctive: line 53
If verifying would take a tool or an account you do not have, say exactly that and tell the user the one thing they can click to settle it.
“Common” is measured: kept by at least 6 in 10 of the 12 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
/reality-check
Work runs on a picture in your head of where everything stands. When that picture is wrong, you chase someone who already replied, you never follow up with someone who is waiting on you, or you tell your team something shipped that is still sitting in a branch. Each one costs something real.
This is the gate that keeps the picture true.
The rule
A tracker is a lead. The live thing is proof.
Notes, spreadsheets, CRM fields, ticket statuses, memory of last Tuesday, and anything the model said earlier in the session are all leads. They were true when written. They decay within days.
Before any of these words reach the user, open the actual artifact:
| Word | What counts as proof |
| sent | The message in the Sent folder, or the thread itself. Not a draft |
| replied / heard back | The reply, opened and read |
| no response / ghosted | The thread itself, showing nothing after the last outgoing message, and the date |
| scheduled | The event on the calendar |
| shipped / deployed / live | The change visible in production, or the deploy log showing it succeeded |
| merged | The PR showing merged, not approved |
| done / closed | The thing itself in its finished state, not the ticket status |
| followed up | The follow-up message in the thread |
How to run it
- List every checkable claim you are about to make. Anything with a status word in it.
- Open the live source for each one. The thread, the calendar, the PR, the dashboard, the production page. Not a search result, the thing itself.
- Correct what turns out to be wrong, and say plainly that it changed. "The tracker said sent. The thread shows a draft. It never went out."
- Label anything you could not verify. Say which source you could not reach and why. Do not quietly drop it and do not guess around it.
Then deliver the list.
Three traps
Absence is not evidence. An empty search does not prove nothing was sent. Search hides things: wrong account, archived, a different address, filtered to a label. Open the thread itself, or say "I could not find it, which is not the same as it not existing."
A send is not a receipt. Sent proves you sent. It does not prove they read it, and it never proves they will answer. The same goes for a merged PR: merged is not deployed.
Old is not gone. Something that went quiet five weeks ago is not automatically dead. Check its current state before writing it off, because "still open, no movement" and "closed" call for completely different moves.
The honesty floor
- If you did not check it, say you did not check it. In the same sentence as the claim, not in a note at the bottom.
- Never say something is done because it should be done by now.
- Never infer that someone replied, or that something shipped, because time has passed.
- If verifying would take a tool or an account you do not have, say exactly that and tell the user the one thing they can click to settle it.
A short verified list beats a long confident one. Every time.
Pairs with no-bullshit: that one governs factual claims, this one governs status claims.
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.