Paper Trade Policy
Checks a simulated buy/sell setup against a predefined strategy and current supplied evidence.
Choose a Classifier to read its AI Agent Instructions, outputs and examples.
Checks a simulated buy/sell setup against a predefined strategy and current supplied evidence.
Flags replies that break character or world facts, or take control of the player.
Marks context to preserve exactly, summarize, or omit before an agent continues.
Separates rent, deposit and lease dates from other amounts and dates.
Screens whether two business references may refer to the same organization, not the same person.
Flags messages against stated community rules without treating all criticism as abuse.
Flags changed payment instructions for independent verification.
Checks whether a proposed subgoal duplicates work already assigned or completed.
Separates review sentiment, complaints, feature requests and stated repurchase intent.
Applies a saved feed preference without rewriting the rule for every post.
Identifies whether a passage states a duty, permission, condition or background fact.
Selects an explicitly supplied tool option rather than inventing a tool call.
Flags suspicious link evidence for review without claiming a URL is safe.
Checks whether observed browser results satisfy the requested goal and constraints.
Labels the stance expressed about an asset in supplied posts; it does not forecast returns.
Separates blocking review findings from advice while parallel checks finish.
Loads relevant conditional guidance without dropping mandatory repository rules.
Finds posts centered on a provocative opinion rather than an ordinary update.
Rates reported operational urgency separately from how angry the customer sounds.
Classifies a search query’s intent relative to the advertised product before keyword review.
Checks whether a support reply addresses the request without inventing a resolution or commitment.
Distinguishes paid-promotion language from the main content of a transcript.
Maps a visual brief to predefined design tokens rather than generating application code.
Routes code-change descriptions toward the right review depth; it does not certify code as secure.