dancing-bear: Instructions file for Gemini CLI

Gemini Familiarize

Minimal-token guidance for Google Gemini to understand this repository.

Startup: Automated Capsule

Generate and ingest the summary capsule:

./bin/llm familiar --stdout

This yields a YAML snapshot covering policies, roadmap, migration status, and agent workflows.

On-Demand Context

Load only as needed:

• .llm/CONTEXT.md — architecture, rules, commands.
• ./bin/llm domain-map --stdout — directory/auth map (generated on demand, not a file).
• .llm/PATTERNS.md — code templates for filters, CLI wiring, tests.

Tool Discovery: Use --agentic

Prefer agentic schemas over --help:

./bin/mail --agentic --agentic-format yaml --agentic-compact
./bin/calendar --agentic --agentic-format yaml --agentic-compact
./bin/schedule --agentic --agentic-format yaml --agentic-compact
./bin/llm agentic --stdout

All 19 apps support --agentic. ./bin/llm inventory --stdout prints the exact invocation for each — ./bin/<app> is wrong for three of them.

Token Budgets

Respect token budgets defined in .llm/AGENTIC_BUDGETS.yaml.

Quick Reference

• Primary CLIs: ./bin/mail, ./bin/calendar, ./bin/schedule, ./bin/phone.
• Auth: --profile gmail_personal|outlook_personal (paths in ~/.config/credentials.ini).
• Tests: make test (preferred); never use bare python3 -m unittest in a worktree.
• Config SoT: filter YAML lives in config/; generated output goes outside the checkout (src/core/paths.py — --out-dir, $DANCING_BEAR_DATA_HOME, $XDG_DATA_HOME/dancing-bear, ~/.local/share/dancing-bear).