remediation

Traces and repairs downstream dependents of a false claim with expiring currency verdicts.

10|2|Updated Jan 24, 2026
One-click install
npx skills add https://github.com/nrdxp/predicate --skill remediation-nrdxp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: remediation
Source: https://github.com/nrdxp/predicate/tree/main/skills/remediation
Command: npx skills add https://github.com/nrdxp/predicate --skill remediation-nrdxp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a claim, citation, or premise already relied upon turns out to be false, detection alone does not fix the damage — downstream artifacts built on it remain wrong. This Skill provides a disciplined Select → Produce → Accept procedure for enumerating everything that depends on the false claim, repairing each dependent correctly, and closing with a scoped, expiring verdict instead of an unqualified "done." ## Core Features & Use Cases - Dependent enumeration (SELECT): Uses the corpus dependents_of query rather than grep to list every artifact resting on the false claim, with external dependents surfaced as residuals. - Species-routed repair (PRODUCE): Classifies claims as born-false or became-false and dependents as composed or hand-authored, routing each to the correct repair — regenerate, re-establish, supersede, or halt for a human closer on closed-claim impeachment collisions. - Expiring acceptance (ACCEPT): Closes only with a currency verdict {scope, anchor} whose scope is confirmed against the dependent set by a second party, since "clean" is a theorem-bounded expiring claim, never a standing fact. - Use Case: An agent discovers a test invariant was derived from a since-refuted premise; it invokes this Skill to enumerate all dependents via the corpus, repair each by its classification, and record a scoped currency verdict with named honest limits. ## Quick Start Invoke the remediation skill to trace everything that depends on the false claim I just found, repair each dependent, and close with a scoped currency verdict.

Frequently Asked Questions about remediation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fix downstream artifacts after a claim turns out to be false?

Follow the Select → Produce → Accept procedure: enumerate dependents with the corpus dependents_of query, classify each as composed or hand-authored, repair accordingly, then close with a scoped currency verdict rather than an unqualified done statement.

What is the difference between born-false and became-false claims?

A born-false claim was never true at its own anchor, while a became-false claim was true when written but falsified by later record growth. Species is derived from anchor comparison, never authored, and determines which repair path applies.

Why use dependents_of instead of grep to find affected artifacts?

Grep answers questions about text while the corpus answers about extracted entries, and the two disagree. The dependents_of query walks derivation edges like because and depends, which text search cannot reconstruct.

Can an agent auto-resolve a false claim that was already closed?

No. A born-false refutation against a closed claim is a proven-contradictory collision that the model cannot adjudicate internally. The walk must author an anchored refutes edge and halt for a human closer, never auto-resolve.

What are the limitations of this remediation procedure?

Semantic re-derivation of a claim in entirely different words is not detected, scope adequacy for currency verdicts is tracked as unsolved tech debt, and species routing plus currency gates depend on the unmerged feat/refutation-and-currency branch with manual fallbacks.