| name | nano-core |
| description | Understand and operate nano-core itself. Use for questions about nano-core architecture, behavior, configuration, capabilities, skills, installation, troubleshooting, Telegram administration, or coding delegation; route focused work to the matching nano-core skill. |
nano-core
Use this as the self-referential overview and router for nano-core. Read current repository files and inspect runtime state before making factual claims; the skill is guidance, not a substitute for evidence.
When to use this skill
- Use when the user asks how nano-core works, what it can do, or how it is configured.
- Use first for nano-core self-help, then load the focused skill when the request crosses one of the boundaries below.
When not to use this skill
- Do not use for ordinary domain work merely because nano-core is the host.
- Do not keep using this overview when a focused skill below covers the request.
Route focused requests
- Installation, onboarding, upgrades, prerequisites, or service setup:
nano-core-setup.
- Failures, logs, provider errors, routing incidents, locks, or runtime diagnosis:
nano-core-debug.
- Telegram main-chat claims, command scope, delivery mode, media, or routing:
nano-core-telegram-ops.
/coder, /coding, plan-only delegation, approval behavior, or delegated-run progress: nano-core-coder-ops.
Use skill_view to load the selected skill before following its workflow. A request may require more than one focused skill, but load only those that materially apply.
Architecture facts to verify
- nano-core is a Node.js host process that receives chat messages, persists them in SQLite, runs the agent (Pi) inside an isolated container per chat group, and returns the response to the originating chat.
- Domain behavior is added through capability profiles, not baked into the core;
core is the default and always available.
- Project runtime skills (
skills/runtime/) and main-workspace skills are merged into a per-group managed skills directory before each run. Pi receives the same set through its configured agent skill directory.
- Main/admin chat has privileged controls (coder delegation, service/gateway operations, task and group management). Non-main chat behavior is intentionally narrower and gated by the trigger word.
MEMORY.md is durable memory; SOUL.md is stable identity and policy, not a compaction log.
Verify these facts against README.md, AGENTS.md, CLAUDE.md, docs/, configuration, and runtime code when those files are available.
Guardrails
- Never run destructive git commands unless explicitly requested.
- Preserve unrelated worktree changes.
- Privileged administration and coding delegation are main/admin chat only.
- Never expose credentials or paste secret values into chat, logs, commits, or diagnostic output.
- Distinguish portable Agent Skills fields (
name and description) from nano-core-specific policy in this body and skills/manifest.json.
