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.
Project research columns with an explicit keep list
Create a minimal working table while retaining a traceable mapping to authorized input fields.
Read guideMake a research table sort reproducibly
Specify tie breakers and missing-value rules so identical inputs produce an explainable row order.
Read guideExpand nested evidence arrays without inflating parent counts
Keep parent and child identities explicit when flattening returned arrays into rows.
Read guideValidate table types without coercing away evidence problems
Reject unsafe conversions and retain raw values when normalizing returned research fields.
Read guideMake a cleanup step safe to apply more than once
Test that repeated local normalization does not repeatedly alter already-normalized evidence.
Read guideMinimize fields in a permission-aware research handoff
Build a recipient-specific projection instead of forwarding a complete workspace export.
Read guideRecheck access before redistributing a saved table
Treat current authorization as a release gate rather than inheriting an old export's permissions.
Read guideValidate redaction across table and download views
Check that a minimized artifact does not leak withheld fields through secondary representations.
Read guideAudit evidence links before a team handoff
Separate original public references, internal artifact paths and access-bound download links.
Read guideHandle a denied research handoff without bypassing policy
Keep denied recipients blocked while preserving a minimal operational record of the issue.
Read guideCreate a review queue with actionable operational states
Use issue type, artifact revision and closure evidence instead of vague review flags.
Read guidePrioritize operational review without hiding older blockers
Rank queue items by release dependency and age while retaining unresolved access problems.
Read guideSeparate preparer and checker for a table release
Require an independent operational check of membership and schema before handoff.
Read guideRoute conflicting operational reviews without overwriting either result
Preserve checker disagreement and assign a specific resolution question.
Read guideReopen reviewed rows only when a relevant dependency changes
Use lineage-based triggers to avoid both stale approvals and unnecessary full rechecks.
Read guideTriage a partial response without discarding accepted evidence
Separate valid returned rows, pending coverage and actual failures in an operational recovery ledger.
Read guidePause and resume evidence operations after a rate limit
Honor returned retry guidance while preserving the exact last committed operation.
Read guideResolve timeout ambiguity before repeating a table operation
Distinguish no response, received output and committed local state after an interruption.
Read guideQuarantine a malformed result payload without losing its context
Keep parsing failures separate from empty evidence and preserve a minimal diagnostic reference.
Read guideVerify operational recovery without overstating evidence completeness
Check a resumed service against the affected table and its remaining coverage limitations.
Read guideEnforce one-to-one joins between research table revisions
Reject duplicate-key matches before reconciling fields from two authorized outputs.
Read guideValidate many-to-one enrichment without multiplying source rows
Check lookup uniqueness before adding permitted context fields to evidence records.
Read guideMake many-to-many expansion explicit in evidence tables
Represent link relationships without presenting expanded rows as additional independent sources.
Read guideKeep unmatched research rows in an operational ledger
Preserve left-only, right-only and invalid-key states instead of silently losing rows during joins.
Read guide