# General-purpose research framework: completed intake disposition

Date: 2026-09-08. Source: A General-Purpose Deep Research Framework.md,
40,346 bytes, SHA-256 bd056712bf9bd0f2dad7b09ce698150fb5ccb1c228f5438c98824dc36abe87de.
Disposition: incorporated into the existing internal research-review owner.

The complete file was read: executive framing and unspecified parameters;
question trees, claim tests and eligibility; source architecture, search layers
and data acquisition; provenance and permission separations; quality checks;
qualitative, statistical, causal, predictive and mixed methods; evidence/report
templates, graphics, diagnostics, examples, timelines, budgets, bibliography
and final seven-step plan. No missing attachment is represented as read.

## Selected use and actual owner

The current user wants a completed improvement intake, not a newly commissioned
domain study or a research platform. .uai/research-review.uai already implements
decision scope, complete-read identity, claim/evidence separation, contrary
evidence, exact locators, bounded acceptance and rejection, implementation and
validation. This source reinforces those existing rules. The owner now expressly
adds raw-versus-normalized custody, transformations, units, missingness, duplicate
source relationships and the distinction between prediction and causation.
These are selected internal working conventions, not claims of certification
under any named external research standard.

This turn exercises those rules: the public glossary parity test exposed source
drift, the source label was reconciled to existing visible wording, and duplicate
machine terms were repaired without declaring external indexing or adoption.
The legal delivery summaries were read but their missing bundles and opaque
citations were not promoted into product facts.

## Not adopted, with reasons

The example healthcare question, causal models, clinical recruitment/IRBs,
commercial databases, field interviews, slides, paid tools and staffing model
are not features requested for this coordination commons. The sample budgets,
dates, source-count targets and staffing durations are illustrative, not
implementation gates. A research lead or human reviewer is not installed as
runtime authority. Confidence labels or statistical diagnostics must not become
participant trust/reputation scores or a cognitive profile.

The supplied chat citation tokens and claims about PRISMA, Cochrane, statutes,
vendor tools, APIs and standards remain leads, not independently checked facts.
No such external claim is published or used to assert compliance. No new crawler,
regulated service, participant experiment, paid purchase or external publication
is authorized or implemented by this source.

There is no accepted unfinished feature being moved to a backlog. The selected
change is the small research-owner clarification above; other suggestions are
explicitly unadopted for this task. Validation consists of that source-to-owner
comparison, current-memory tests and the intake ledger/custody integrity check.
The exact raw source is preserved under reports/input/v049 and excluded from
the public root deployment. Only its verified intake copy may be removed.