v0.1.5

• Added combined prompt topological_fraud for lineage disconnection

v0.1.4

• override llm prompts
• select llm prompts to be executed
• specify folder with CPGs
• improved report view

v0.1.3

• Move CPG spec, structural reference data, and response-format rules into a single json/shared_system.md sent once as a SystemMessage per snapshot (reused across all five Prompt calls, enabling prompt caching), leaving each json/_user.md with just the attack-specific objective.
• Add $ROOT_VARIABLE_IDS$/$LEAF_VARIABLE_IDS$/$MULTIUSED_VARIABLE_IDS$/$DUPLICATE_NAME_LEAF_IDS$ placeholders to the standalone markdown templates too, matching the JSON path.
• Add chronology validation (ProcessingResult.isValidChronology/invalidateChronology) alongside the existing replay-mismatch check.
• Write each run's output under a timestamped /out/ directory instead of a flat out/, so repeated runs don't overwrite each other.
• Clean up CLI logging (custom single-line formatter) and cpgpath parsing order in Main.

v0.1.2

• Improved prompts
• Improved output processing

v0.1.1

WF updates

v0.1.0

Initial release