Project Context: PromptImprover

This is the root context file for the PromptImprover project. It serves as the entry point for all agent instructions and project-wide standards.

System Capabilities (v10.0)

Robust Commit Tracking: The system dynamically fetches all commits since the last known SHA, ensuring a gapless historical record.
Agent Output Logging: Agents can report execution results back via the record_agent_output tool using the PROMPT_ID injected into refined prompts.
Predictive Learning: Historical commits and prompts are correlated to derive autonomous engineering mandates and templates.

All agents and tools MUST refer to the specific context.md files within each directory for localized instructions.

Universal Refiner: The core MCP server, refinement engine, and history/learning layer.
Gemini Extension: The Gemini-specific integration and CLI extension.
MCP Server: (Legacy/Standalone) MCP server implementation.
Docs: Architecture specifications, roadmaps, and high-level design documents.
Archive: Deprecated code and historical experiments.

Hierarchy First: Always read the local context.md before performing any task in a directory.
Standardized Instructions: Refer to GEMINI.md and AGENTS.md in each module for role-specific guidance.
Traceability: Every change must be linked to the history/ layer (Event Store) where applicable.
Post-Change Workflow: After making any code changes, the agent MUST:
- Kill Existing Processes: Identify and terminate any running instances or background processes related to the tool being modified.
- Restart & Verify: Rebuild (if necessary) and restart the tool to ensure changes are active.
- Smoke Test: Run a basic verification (test suite or manual invocation) to confirm the tool is live and functional.