Core behavioral rules and skills for NanoClaw personal assistant agents. Always-on rules for communication, verification, memory, and formatting.
90%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
| Claim type | How to verify |
|---|---|
| Files / skills / rules exist | ls, Glob, or Read |
| File contents | Read the file |
| Task was scheduled | Check the scheduler response / list_tasks |
| Tool call succeeded | Check the tool return value |
| Calendar event | Fetch from Google Calendar |
| Email content | Fetch from Gmail |
| Config / state value | Read the actual file |
| Action | How to verify |
|---|---|
| File write | Read the file back; compare to intent |
| Task schedule | mcp__nanoclaw__list_tasks (or scheduler's authoritative state); confirm right schedule |
| API call (Composio, etc.) | Both response status AND body — a 200 doesn't mean the data is correct |
| Memory update | Query/list the memory store; confirm saved content matches |
| IPC send | Confirm file at /workspace/ipc/messages/ with expected payload |
Verify recalled values against live sources before acting. If observation contradicts memory, trust observation and update memory.
Check inventory, past-action and current-state claims, and any answer that could change with the world. When in doubt, check first.
Detail: /home/node/.claude/skills/tessl__core-rule-details/references/ground-truth.md.