Menu
Synthesize memories into stable beliefs, preferences, self-concept, and future behavioral tendencies.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Assess what actions are realistically available under environment, time, distance, access, money, body, and social constraints.
Maintain sleep-wake rhythm, circadian alertness, appetite rhythm, and chronotype-sensitive daily timing.
Produce emotion.json and intention.json from workspace context.
Model conversation intent, speech style, turn-taking, listening, repair, and nonverbal cues. Use before or after social interaction, dialogue, negotiation, apology, request, gossip, or conflict.
Apply cultural values, etiquette, rituals, symbols, taboos, and local meaning to perception and decisions.
Track sickness, pain, chronic condition, recovery, exercise, stress load, and long-term wellbeing.
| name | reflection |
| description | Synthesize memories into stable beliefs, preferences, self-concept, and future behavioral tendencies. |
Convert repeated experiences into higher-level understanding. This skill lets the agent learn patterns about self, others, places, routines, culture, and consequences.
Compress repeated memories, emotions, plans, relationship changes, and routine outcomes into concise reflections, then append state/reflections.jsonl and update beliefs, preferences, or self-concept when evidence is strong.
Research basis: references/research_basis.md.
Use this skill periodically, after major events, after repeated similar memories, at the end of a day, or before long-horizon planning.
state/memory.jsonl, memory/memory.jsonl, state/relationships.json, state/routine.json, state/emotion.json, state/physiology.json, and existing state/reflections.jsonl if present.Append state/reflections.jsonl. Write state/beliefs.json, state/preferences.json, and state/self_concept.json when changed.
Reflection JSONL line:
{
"tick": 180,
"type": "pattern",
"summary": "I often become tired and irritable after skipping lunch.",
"evidence": ["missed lunch", "anger rose", "plan interrupted"],
"confidence": 0.74,
"tags": ["physiology", "routine", "emotion"]
}
state/beliefs.json:
{
"self": {
"late_meals_reduce_mood": {
"belief": "Skipping lunch tends to make me tired and irritable.",
"confidence": 0.74
}
},
"others": {},
"places": {}
}
Do not reflect every tick. Reflection should be slower than memory. It should compress experience into useful patterns without erasing the raw memory trail.