continuity
par riley-coyote
Notes & PKMRésumé
Asynchronous reflection and memory integration for genuine AI.
Description
Ce que fait ce skill
Runs post-session reflection to extract structured memories categorized by type (fact, preference, commitment, relationship, etc.) and assigned confidence scores. Generates follow-up questions from each session and stores them for the next conversation. When the user returns, surfaces relevant questions for a contextually aware session start.
Unlike passive note-taking or raw session logs, it actively integrates memories and generates questions, so the next session opens with genuine continuity rather than performed familiarity.
Quand l'utiliser
- Resuming a long-running project after days away without losing context
- Tracking commitments an agent made across multiple sessions
- Surfacing unresolved questions at the start of a new conversation
- Maintaining a persistent self-model for an AI agent over time
- Auditing what preferences or facts were learned in prior sessions
Exemple de workflow
User returns after a 2-day break on an ongoing agent infrastructure project
- 1
Heartbeat triggers
continuity reflectafter session idle exceeds threshold - 2
Analyzes the previous conversation and extracts typed memories with confidence scores
- 3
Generates follow-up questions for unresolved threads and stores them to questions.md
- 4
Updates identity.md with any growth or narrative changes noted in the session
- 5
Runs
continuity greetat session start to surface the top pending questions
Agent opens with: "Welcome back. A few things from last time: How is the Lovable backend progressing? Has the token launch timing been decided?"
Installation
clawhub install riley-coyote/continuityExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.