What problem does it solve?
Default claude-mem observation modes are tuned for software work, so users in law, architecture, research, or other domains end up with notes that miss the categories they actually search for later. This Skill interviews the user, proposes a domain-specific taxonomy, installs it durably, and proves the active mode appears in startup context.
Core Features & Use Cases
- Interactive taxonomy design: Conducts a purpose-first interview, then proposes 4–8 observation types and 4–8 concept tags tailored to the user's domain.
- Validated installation: Merges the override with a parent mode, validates every required prompt field, and writes the mode file atomically under the resolved data directory.
- Optional Telegram alerts: Securely configures a Telegram bot token via hidden terminal input, validates it with getMe, discovers the chat ID, and merges trigger types and concepts without overwriting existing settings.
- Restart and verification: Restarts the worker through the verified CLI path and confirms the active mode appears in startup context, falling back to repair if the worker reverts to code mode.
- Use Case: A law student wants claude-mem to remember case holdings, issue-spotting patterns, professor frameworks, minority rules, and exam traps. This Skill interviews them, proposes a focused remix of the bundled law-study mode, installs it, and verifies the active mode in startup context.
Quick Start
Ask the mode-creator skill to set up a custom claude-mem mode for tracking architecture design decisions, client priorities, and site discoveries, with Telegram alerts for cost-impact concepts.
