proof-standard
Skill73 starsChanged 3 months agoMITView on GitHub
Never make an external-facing claim about a product without evidence from a truth file. Use when writing marketing copy, announcements, docs, landing pages, investor material, or answering "can our product do X".
In this repository15 other files it gives its agents.
- brain-initSkill
- checkpointSkill
- decision-ledgerSkill
- doctorSkill
- goalSkill
- knowledge-baseSkill
- pinSkill
- recallSkill
- rehearsalSkill
- reportSkill
- researcherSkill
- scheduleSkill
- setupSkill
- skill-forgeSkill
- verify-gateSkill
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 35
Truth file first: Before any capability claim, open the truth file.
- Distinctive: line 37
Product changes → truth file first, copy second: Never the reverse order.
≈ “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
Proof standard
External claims about a product come from evidence, not from memory, enthusiasm, or the roadmap. The failure mode this prevents: shipping confident copy about features that don't exist.
The truth file
memory/PRODUCT-TRUTH.md (run brain-init to create it) has three sections:
- Implemented — with evidence (code path, test, screenshot) and a date
- Not implemented — explicitly listed, so absence is a fact rather than a gap
- Permanently excluded — decided against; copy must never imply it
Claims may be sourced only from Implemented.
One truth file per subject you make claims about, split by what kind of evidence is available — code you can read versus someone else's system or word. Name the evidence class at the top of each file, and never let two classes share a section: a line you can settle by opening a path and a line that rests on a vendor's statement do not carry the same weight, and a shared section hides which one you are holding.
Four code states — never blended
| State | Meaning |
implemented | The code exists |
wired | Connected end-to-end (UI → backend → effect) |
operational | Running in production with real data |
verified | Tested, with evidence you can point to |
Copy that says "we do X" requires operational or better. "X is available" requires at least wired. Everything else is roadmap language — clearly future-tense, clearly conditional. Mixing these states in one sentence is how honest teams end up lying.
Rules
- Truth file first. Before any capability claim, open the truth file. No entry → no claim; register an open question instead of guessing.
- Evidence and dates on every entry. "It works" is not an entry. "Import wired end-to-end, e2e test passing, 2026-08-01,
tests/import.spec.ts" is. - Product changes → truth file first, copy second. Never the reverse order.
- Internal is external. The deck you show an investor, the answer you give a partner — same standard. There is no audience for unproven claims.
- Absence claims too. "We never store X" is a capability claim about the negative — it also needs a truth-file entry with evidence.
Freshness
Every entry carries the date it was last checked against the code, not only the date it was written. Before an entry older than 90 days is used in a claim, re-confirm it — open the path, run the test, look at the screen — and stamp the new date. A date says when the entry was true, not that it still is.
"Not implemented" is the section that leaks the other way. A stale Implemented entry pushes you toward claiming too much, and everyone watches that direction. A stale "Not implemented" line does the reverse: the capability shipped weeks ago, the line never moved, and the truth file is now talking you out of a claim you have already earned — understating instead of overstating. Sweep that section on the same 90-day clock, and check it again every time something ships.
Part of ballast — foundations first, every solved path kept, done means checks pass.
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.

