Pattern Index

Generated from catalog/patterns.json.

ID	Name	Summary
pattern-direct-instruction-framing	Direct instruction framing	State the task, audience, and success condition in the first lines.
pattern-context-boundary-marking	Context boundary marking	Separate instructions, trusted context, untrusted context, and input data with clear labels.
pattern-role-responsibility-framing	Role and responsibility framing	Define the assistant's job, authority, and refusal boundary without adding theatrical personas.
pattern-output-contract-first	Output contract first	Specify format, fields, ordering, and validation rules before examples or data.
pattern-few-shot-calibration	Few-shot calibration	Use compact examples to calibrate tone, structure, and edge-case handling.
pattern-positive-constraints	Positive constraint rewrite	Replace broad negative rules with concrete desired behavior and allowed alternatives.
pattern-decomposition-planning	Decomposition planning	Ask the model to break complex work into auditable stages before execution.
pattern-checkpointed-execution	Checkpointed execution	Add stopping points for review when the task changes files, spends money, or uses tools.
pattern-retrieval-grounded-answering	Retrieval-grounded answering	Force answers to use supplied sources and label unsupported claims.
pattern-citation-before-answer	Citation before answer	Have the model collect evidence snippets before synthesizing a response.
pattern-clarify-or-proceed-policy	Clarify-or-proceed policy	Define when to ask a question and when to continue with stated assumptions.
pattern-assumption-ledger	Assumption ledger	Record uncertain assumptions separately from verified facts.
pattern-decision-matrix	Decision matrix	Compare options with criteria, weights, evidence strength, and tradeoffs.
pattern-rubric-driven-critique	Rubric-driven critique	Evaluate outputs against a named rubric before revising.
pattern-diff-explained-rewrite	Diff-explained rewrite	When rewriting, show what changed and why without restating the whole source.
pattern-acceptance-self-check	Acceptance self-check	Make the model verify final output against explicit acceptance criteria.
pattern-tool-selection-policy	Tool selection policy	Define which tools are allowed, when they are required, and when approval is needed.
pattern-tool-result-provenance	Tool result provenance	Keep tool outputs traceable to source, timestamp, and transformation.
pattern-long-document-map-reduce	Long-document map-reduce	Process large inputs in chunks, then merge with conflict handling.
pattern-state-handoff-summary	State handoff summary	Create compact state summaries that preserve decisions, open items, and evidence.
pattern-delegation-contract	Multi-agent delegation contract	Give subagents bounded goals, inputs, outputs, and review requirements.
pattern-safe-refusal-alternative	Safe refusal with allowed alternative	Refuse disallowed requests while offering a useful safe path.
pattern-schema-bound-extraction	Schema-bound extraction	Extract data only into a declared schema with null and uncertainty rules.
pattern-error-recovery-prompt	Error recovery prompt	Recover from failed outputs by identifying the violated contract and retrying narrowly.
pattern-prompt-compression	Prompt compression	Compress long instructions while preserving priority, constraints, and evidence.
pattern-eval-case-generation	Eval case generation	Generate representative, adversarial, and regression test cases from success criteria.