percept-ambient
par jarvis563
AI & LLMsRésumé
Ambient intelligence mode — continuous context awareness without explicit commands.
Description
Ce que fait ce skill
Runs continuously in the background, capturing ambient conversations and building a local knowledge graph of people, projects, and relationships. When an agent needs context, it assembles a structured Context Packet from accumulated history — no explicit commands required. All data is stored locally with configurable retention and privacy controls.
Passive accumulation means agents have situational awareness without users manually logging notes or issuing commands.
Quand l'utiliser
- Recalling who was involved in a past project decision
- Answering 'what do you know about this person?' without manual notes
- Giving an agent background context before starting a task
- Tracking which projects are active across daily conversations
- Searching conversation history semantically for a specific topic
Exemple de workflow
User asks: What do you know about the Johnson account?
- 1
Receives natural-language query
- 2
Runs hybrid FTS5 and vector search over stored utterances
- 3
Resolves entities and relationships linked to 'Johnson account'
- 4
Assembles a Context Packet with recent conversations, entity data, and relationship mappings
- 5
Returns structured context to the requesting agent
Context Packet containing recent conversations mentioning the Johnson account, resolved entity records, and mapped relationships
Installation
clawhub install jarvis563/percept-ambientExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.