FORTRESS Truth: A Linter That Blocks Unsourced Claims

"Trusted by thousands" is a sentence shaped like a fact. If nothing backs it, publishing it is a risk you took without noticing. This installs a claim linter and a register that decides, per claim, whether you can say it.

The register carries five states: Cleared, Uncleared, Changed, Contradicted, and a canonical source. Section 2.1 explains which direction each state is read, which matters because a Changed claim and a Contradicted one call for opposite actions. Staleness is handled explicitly, so a claim cleared in March doesn't stay cleared forever by default.

It ships three working files: claim_lint.py, a config, and a register template. Section 3 names the banned move, section 4 covers what gets installed, and section 5 covers running it, so this is a tool you execute rather than a policy you remember.

At 4,076 words it's the longest skill in FORTRESS, and it's the one the drafting skills depend on.

This is one of the seven FORTRESS skills. It reads and writes the shared .monkeys/ pack, so it works on its own and works better alongside the rest.

Where this came from

Deliberon is a Windows desktop app that runs a council of AI agents on a hard decision and hands back a Decision Record: dissent preserved, every claim marked as proved, asserted, or estimated.

Truth is the linter the rest of the system writes against, and the reason a draft can be checked instead of trusted.

Pay once, it's yours, no subscription. Thirty days of the full council with no account and no card. Runs on your machine with your own model keys.

https://deliberon.com

The rest of my work, organised by material: https://ryan-heltemes.com