v0.1.2 - Reliability & Security Hardening

🚀 Summary

Critical infrastructure hardening and security sandboxing to prepare the system for safe, high-concurrency production use.

✨ Key Features & Improvements

🛡 Security Sandboxing

• What changed: Implemented a `run_safe_command` utility with a strict executable allowlist and hardened path traversal validation.
• Reasoning: Since the AI can generate shell commands, we must ensure it cannot execute dangerous scripts or access files outside the project directory.

🔒 Data Integrity & Atomicity

• What changed: Added `filelock` and temporary file writes for knowledge base updates, along with PII/Secret scrubbing.
• Reasoning: Prevented data corruption during parallel agent runs (race conditions) and ensured sensitive keys are never leaked into the vector database.

📐 Standardized Utility Architecture

• What changed: Reorganized the flat `utils/` folder into categorized sub-packages (`io/`, `security/`, `knowledge/`).
• Reasoning: As the project grows, a flat structure leads to circular imports and "god objects." This modularity makes the codebase easier for both humans and AI to navigate.

🐛 Issues Addressed

• #42 [Security]: Sandbox isolation for command execution.
• #43 [Data Integrity]: Atomic indexing and PII protection.
• #48 [Architecture]: Standardization of logging and utility interfaces.

📝 Technical Notes

• Standardized logging to use a centralized `rich.console` provider in `utils/io/logger.py`.
• Updated release-tagging permissions (CI/CD fix).