Public methodology · version 1.0

From source to decision.
Every boundary stays visible.

Agenova publishes Agent profiles, Source-Backed facts, Decision Briefs, explicit evidence gaps, accepted changes, and dated snapshots. It does not publish a universal score, ranking, winner, or purchasing recommendation.

Effective 2026-08-25Last updated 2026-08-25Applies to public Agent facts and Decision BriefsStatus current
01 · Current public product

What Agenova publishes now

Official sourceTraceable, scoped evidence
→
Accepted claimHuman- or policy-reviewed
→
Agent factEvidence state disclosed
→
Decision cellCovered or explicit gap
→
Dated snapshotChange remains auditable
02 · One public vocabulary

Evidence states

These labels are the public contract used across profiles, comparisons, and Decision Briefs.

Unverified

A provider-listed or discovered answer without accepted evidence that directly supports the claim. It does not mean the capability is absent.

Source-Backed

A claim supported by accepted, traceable evidence whose source and scope directly match that claim.

Execution-Verified

A claim supported by a publication-approved, versioned execution result. Official documentation alone cannot create this state.

Intelligence gap

No accepted answer exists for the selected candidate-requirement cell, or the available evidence is partial, stale, or outside the claim scope.

System-qualified identity

Automated identity rules found a likely canonical relationship. This is an internal qualification signal, not a reviewed product fact.

03 · Claim and source rules

A source never proves more than it says

01

First-party first

Maintainer-controlled product pages, documentation, pricing pages, policies, release notes, and official repositories may support claims within their stated scope.

02

Claim-level scope

A source supports only what it directly states. A security page cannot prove pricing; a product page cannot prove independent execution reliability.

03

Partial evidence stays partial

A source that answers only part of a field cannot be expanded into a complete claim.

04

No negative inference

Not finding documentation is recorded as Unverified or an Intelligence gap, never as proof that a feature is unavailable.

05

Third-party boundary

Third-party reporting can guide research, but does not by itself create a Source-Backed product claim.

06

Reachability is not truth

A successful HTTP check confirms source access, not the truth or completeness of every claim on that page.

04 · Decision methodology

Evidence coverage is a cell count—not a product score

Each selected candidate × selected requirement creates one equally weighted decision cell. A cell counts as covered only when its current answer is Source-Backed or Execution-Verified.

covered cells÷all selected candidate-requirement cells× 100

Unverified, partial, stale, missing, and unsupported answers remain gaps. Adding or removing a candidate or requirement changes the denominator. Coverage describes the completeness of evidence for this decision; it does not measure product quality.

05 · Decision Brief lifecycle

How a Decision Brief is formed

  1. Frame a concrete question.The audience, workflow, and choice boundary must be explicit. A framework idea is not presented as observed customer demand.
  2. Publish a candidate inclusion rule.The cohort is bounded by category, maintained product identity, public first-party documentation, and evidence needed for the listed requirements. It is not an exhaustive market list.
  3. Select decision requirements.Requirements must be comparable and capable of changing the choice; each is disclosed on the page.
  4. Project accepted evidence into cells.Every answer retains its evidence state and source. Unknowns stay visible.
  5. Apply readiness language.75% or more is “Shortlist ready”; 50–74% is “Directional evidence”; below 50% is “Research first.” None is a recommendation or final purchase gate.
  6. Version material changes.An accepted fact that changes a decision field creates a traceable impact and a new immutable decision snapshot.
06 · Freshness

Current, attention, and source checks

Decision snapshots use the oldest source check in the Pack. “Current” means every verified claim has a dated source check and the oldest check is no more than 35 days old. “Attention” means that window has been exceeded. Missing, invalid, future-dated, failed, changed, or review-required checks cannot silently remain current.

Source monitoring may use a tighter cadence for volatile fields. A reachable URL does not automatically preserve a claim when its content changes.

07 · Accepted change

How a change reaches a decision

Detected source change→ scoped review → accept, reject, or quarantine → versioned Agent fact → affected decision field → new snapshot

Material changes are reviewed against identity, scope, freshness, duplication, and publication rules. Idempotent identifiers prevent the same accepted change from being applied twice; prior snapshots remain historical records.

08 · Publication governance

Concrete publication gates

Source-Backed fact

Canonical Agent identity; explicit claim scope; accepted matching source; review date; valid freshness; publication approval.

Execution-Verified fact

Fixed Agent version; versioned task; recorded environment; atomic observations; evidence manifest; reproducibility review; publication approval.

Decision Brief

Concrete question; bounded candidate rule; explicit requirements; traceable cell states; dated snapshot; visible gaps; no automatic winner.

Foundation Release

Execution publication boundary

Governed execution evaluation is a publication path, not a promise that results already exist. A Source-Backed claim is never relabeled as Execution-Verified. If a suite, environment, Agent version, evidence manifest, reproducibility review, or publication approval is missing, the execution cell remains a gap.

09 · Limits and corrections

What these records do not claim