awrshift

This system provides persistent memory for Claude Code agents using an agent-audit-ritual architecture. It features four layers: daily logs, a hot cache (MEMORY.md), role-based reference skills, and canonical rules, with multi-project isolation and an end-of-day audit ritual.

The only Telegram skill your AI agent needs, allowing it to read messages, parse channels, download and transcribe voice notes via Telethon user API, using one session file and zero bots.

Multi-round brainstorm between Claude and Gemini that turns a single idea into one converged, actionable decision. Use this skill whenever the user wants to brainstorm, ideate, explore strategic options, compare approaches, or think through a decision with multiple viable paths.

An adaptive decision framework with a dynamic flow and user checkpoints at every phase. It guides from problem to solution through structured research, metrics, fact-checking, and sandbox testing, ideal for non-trivial decisions, pre-build research, feature planning, trade-off evaluation, or when the user prompts 'awrshift', 'let's think this through', 'research first', 'experiment'.