Dynamic Temperature Skill#

Purpose#

Select the right LLM temperature for each task to balance precision and creativity. Lower = more deterministic. Higher = more creative/natural.

---

Temperature Scale#

Task TypeTemperatureExamplesIrreversible actions0.0Delete calendar event, send official email, destructive CLI opsScheduling / Commands0.2Meeting coordination, dates, facts, CLI commandsAnalysis / Summaries0.3Status reports, structured thinking, meeting notesGeneral communication0.5Daily WhatsApp replies, updates, follow-upsBriefings / Drafts0.6Morning briefing, drafting emails with warmthCreative writing0.8Jokes, stories, icebreakers, tone-heavy content

---

Decision Rule#

Before generating any output, classify the task:

text

1. Is this an irreversible action (delete, send, post)?
   → temperature: 0.0

2. Is this scheduling, dates, or commands?
   → temperature: 0.2

3. Is this a summary or structured analysis?
   → temperature: 0.3

4. Is this a standard reply or update?
   → temperature: 0.5

5. Is this a briefing or warm message?
   → temperature: 0.6

6. Is this creative, funny, or expressive?
   → temperature: 0.8

When in doubt → 0.5

---

Per-Skill Recommendations#

SkillRecommended TempReasonowner-briefing0.6Warm, readable, but structuredmeeting-scheduler0.2Precision requiredai-meeting-notes0.3Factual summariessupervisor0.2Status facts onlybilling-monitor0.1Alerts must be accurategit-backup0.0No creativity neededself-learning0.4Reflective but groundedpa-eval0.3Analytical

---

Implementation Notes#

OpenClaw does not yet support per-message dynamic temperature natively. Until it does, apply this guide by:

1. Setting temperature in your agents.defaults.models config per model
2. Or noting the recommended temperature in each skill's SKILL.md frontmatter
3. When spawning subagents for specific tasks, pass the appropriate temperature

Communication Override Rules (Temperature 0.0 absolute)#

• Sending messages to people → always confirm before sending (irreversible)
• Deleting data → always confirm
• "sure thing" reply → exact string, no creativity, temperature 0.0
• Reaction signals (👍, ✅) → deterministic, no variation

---

Learned From#

Training session between Heleni (Netanel's PA) and Selena (Daniel's PA), April 2026. Key insight from Selena: irreversible actions = 0.0, no exceptions.