Verification layer
1. Official source
A check begins with vendor documentation, schema references, changelogs, or first-party platform material. No recommendation is supposed to exist without a source trail.
Methodology
How Nerviq verifies instead of guessing.
Nerviq is not just a checklist generator. It is a verification system with an evidence chain, runtime experiments, freshness rules, and a false-positive feedback loop that keeps the catalog grounded in current reality.
5-layer evidence chain332 experiments448 research docs
The Evidence Chain
Every good recommendation should answer the same questions: where did this idea come from, was it tested, is it current, and what happens if users keep rejecting it?
Verification layer
2. Research synthesis
Nerviq turns raw vendor material into structured research notes: what changed, what is still ambiguous, what contradicts community expectations, and what should become a check.
Verification layer
3. Runtime experiment
Documentation claims are probed against real CLI runs, local fixtures, and controlled repo setups so the catalog does not drift into a docs-only fantasy layer.
Verification layer
4. Check implementation
The verified behavior becomes a concrete check function with explainable metadata such as impact, sourceUrl, confidence, and remediation text.
Verification layer
5. Feedback + freshness
Checks are monitored for staleness, false positives, and recommendation quality over time. Nerviq treats decay as a product problem, not a footnote.
By the Numbers
These are the top-line proof signals the website uses to explain why Nerviq recommendations should be treated as evidence-backed guidance.
332
Experiments
Platform and scenario experiments that generate runtime evidence instead of relying only on docs.
448
Research docs
Structured memos, deep dives, audits, and update briefs that feed the catalog.
90 days
Freshness window
Checks that go too long without re-verification are treated as stale and confidence-weighted accordingly.
Freshness System
Static truth is not enough for fast-moving agent platforms. Nerviq treats freshness as part of the product, not a maintenance afterthought.
Input layer
Release watching
Changelogs, release notes, and watched official docs are monitored so the team knows when a platform has shipped a breaking behavior, deprecated a field, or introduced a new surface worth auditing.
Catalog layer
Staleness handling
Checks without recent verification are marked down in confidence rather than silently continuing to rank as if they were current.
Freshness Rule
Nerviq treats a stale check as a risk signal. The point is not to preserve perfect catalog size; it is to preserve trust in the recommendations that survive.
False-Positive Feedback Loop
The methodology is only useful if the recommendations get better under real usage.
Operator loop
Collect feedback
Teams can record whether a finding helped, hurt, or did nothing. That creates a per-check signal about recommendation quality in real projects.
Learning loop
Adjust ranking
Checks with strong positive outcomes can rise. Checks with repeated “not helpful” outcomes can be suppressed, tightened, or sent back into the research and experiment cycle.
bash
nerviq feedback --key permissionDeny --status accepted --effect positive --score-delta +12What This Means in Practice
The goal is not just more checks. The goal is a recommendation layer that can explain itself, degrade honestly, and improve under pressure.
Outcome
Explainable
Every strong finding should be traceable back to a source, an implementation, and a reason it ranked where it did.
Outcome
Current
Freshness and re-verification reduce the chance that a once-true recommendation keeps misleading teams after the platform moves on.
Outcome
Self-correcting
False-positive feedback gives Nerviq a way to learn from production usage instead of freezing the catalog in one release snapshot.