One connection. A real first result.
Add the server, sign in and approve, then verify your workspace connection. No key to copy for an OAuth-capable agent.
Compare CSV and JSON without conflating their representations
Check semantic parity across two export formats while preserving type and null distinctions.
Read guideDetect an incomplete export file before distribution
Use parse, envelope and membership checks to identify interrupted or partial result downloads.
Read guidePackage a multi-file research export with a manifest
Connect related files through a minimal manifest without leaking evidence into archive names or indexes.
Read guideCatch unexpected field changes in research exports
Validate exported column names and meanings without silently discarding new or renamed fields.
Read guideQueue stale evidence for purpose-specific review
Separate evidence age from validity and prioritize only the rows that need a freshness check.
Read guidePropagate a source correction through working tables
Track the effect of a confirmed source correction without rewriting historical observations as if they never existed.
Read guidePreserve table structure when evidence must be withheld
Remove prohibited identity exposure without filling a withdrawn row's slot with another record.
Read guideDistinguish an unreadable source from a deleted one
Use explicit availability states rather than guessing why a source can no longer be inspected.
Read guideKeep changing metrics as observations instead of overwrites
Store metric values with observation times and source context so later differences remain interpretable.
Read guideRequire explicit acceptance of a research handoff
Check artifact revision, access and unresolved work before transferring responsibility for a table.
Read guidePrepare a shift handoff for interrupted evidence work
Preserve the last committed operation and next safe action across an operator change.
Read guideAssign every research exception to a responsible owner
Turn vague unresolved notes into bounded work items with evidence references and stop conditions.
Read guideBuild a minimal context packet for a returned table
Give a collaborator enough scope and lineage to continue without sharing unrelated research history.
Read guideHandle a rejected research handoff without losing lineage
Keep a rejected package and its corrections distinct so recipients can verify the replacement.
Read guideNormalize timestamps without losing their original timezone
Keep raw time strings, offsets and precision alongside UTC comparison values.
Read guideUse half-open windows to prevent duplicate boundary rows
Define start-inclusive and end-exclusive intervals for reproducible joins across adjacent reporting windows.
Read guideHandle repeated and missing clock times in evidence joins
Resolve daylight-saving ambiguity before matching local-time records across tables.
Read guideJoin date-only records without inventing exact event times
Preserve calendar-date precision when evidence lacks an offset-bearing timestamp.
Read guideJoin only the shared time scope of two research tables
Compute an explicit common interval before comparing outputs with different requested windows.
Read guideNamespace result identifiers before joining source tables
Prevent identical native strings from colliding across different source contexts.
Read guideProtect numeric-looking source identifiers from coercion
Keep long, leading-zero and exponent-like identifiers intact through tables and exports.
Read guideHandle missing result identifiers without accidental merges
Separate unkeyed rows and prevent null or empty values from matching each other.
Read guideSeparate stable result keys from changeable display labels
Keep renamed titles and handles from becoming accidental new records or false merges.
Read guideUse restricted crosswalks for safe reviewer row references
Give reviewers operational aliases without treating pseudonyms as anonymity or broader access rights.
Read guide