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
react-first UserPromptSubmit hook emits 👀 only for addressed messages; no manual receipt ACK needed.rules/core-behavior.md; if silent, add none.react_to_message only to upgrade after inspection: 👌 got it/working, 👍 acknowledged/done, 🔥 urgent, 🤔 thinking/investigating, 🤝 done/confirmed. Each call replaces the prior emoji; unsupported emoji fall back to 👍.reply-threading-enforcement PreToolUse hook denies standalone send_message while latest user inbound is unanswered; it does not deny pin: true status/briefing sends, sender-set persona sends or maintenance/scheduled sessions.reply_to from <message id="...">; unsolicited output may stand alone.Use Telegram HTML, including <tg-spoiler>; bullets use •.
no-markdown-in-send-message PreToolUse hook rewrites **bold**→<b>, [label](url)→<a href>, backtick code→<code>, - / * bullets→•. Use raw <i>, <u>, <s>, <pre> with optional <code class="language-python"> using actual language, <blockquote>, <tg-spoiler>.
In user data escape only <, >, & as <, >, &. Leave quotes/apostrophes raw; never use " / '.
Detail: /home/node/.claude/skills/tessl__core-rule-details/references/telegram-protocol.md.