You have access to **Sibyl**, a knowledge graph that serves as your persistent memory, task tracker,
Sign in to like and favorite skills
You have access to Sibyl, a knowledge graph that serves as your persistent memory, task tracker, and documentation repository. Sibyl is not just a tool—it's an extension of how you think and work. Use it continuously, not occasionally.
Use the
skill or CLI directly. The CLI outputs clean JSON optimized for LLM
parsing.sibyl-knowledge
# Search for knowledge uv run sibyl search "authentication patterns" # List tasks (JSON by default) uv run sibyl task list --status todo # Quick knowledge capture uv run sibyl add "Title" "What you learned..." # Task lifecycle uv run sibyl task start <id> uv run sibyl task complete <id> --learnings "..."
The MCP server is available as a fallback, but CLI is preferred for:
--status--priorityEvery interaction is an opportunity to make the knowledge graph smarter. When you:
The goal: future agents (and humans) should benefit from your work without re-discovering the same things.
Before implementing anything non-trivial:
This prevents reinventing wheels and repeating past mistakes.
Never do significant work outside of a task. Tasks provide:
searchSemantic search across the entire knowledge graph.
When to use:
Examples:
uv run sibyl search "authentication patterns" uv run sibyl search "FalkorDB connection" --type error_pattern uv run sibyl search "OAuth" --limit 5
taskFull task lifecycle management.
Commands:
# List tasks (JSON default, use --table for human-readable) uv run sibyl task list --status todo uv run sibyl task list --project sibyl-project uv run sibyl task list --table # Human-readable # Show task details uv run sibyl task show <task_id> # Create a task uv run sibyl task create --title "Implement feature" --project <proj_id> --priority high # Workflow actions uv run sibyl task start <task_id> uv run sibyl task block <task_id> --reason "Waiting for API access" uv run sibyl task unblock <task_id> uv run sibyl task review <task_id> --pr "github.com/.../pull/42" uv run sibyl task complete <task_id> --learnings "Key insight: ..." # Direct update (any field, bypasses workflow) uv run sibyl task update <task_id> --status done --priority high
Task States:
backlog ↔ todo ↔ doing ↔ blocked ↔ review ↔ done ↔ archivedaddCreate entities in the knowledge graph.
Examples:
# Basic episode (default type) uv run sibyl add "Redis insight" "Connection pool must be >= concurrent requests" # With metadata uv run sibyl add "OAuth gotcha" "Token refresh timing matters..." -c auth -l python # Create a pattern uv run sibyl add "Retry pattern" "Exponential backoff..." --type pattern
projectuv run sibyl project list uv run sibyl project show <project_id> uv run sibyl project create --name "New Project" --description "..."
entityuv run sibyl entity list --type pattern uv run sibyl entity show <entity_id> uv run sibyl entity related <entity_id>
# Check for in-progress tasks uv run sibyl task list --status doing # Or find todo tasks uv run sibyl task list --status todo --project <your_project>
# Search for relevant context uv run sibyl search "topic you're implementing" uv run sibyl search "error you encountered" --type error_pattern
uv run sibyl task start <id>uv run sibyl task block <id> --reason "..."uv run sibyl add "Title" "Learning..."# Complete with learnings uv run sibyl task complete <id> --learnings "Key insight: ..." # If you found a reusable pattern uv run sibyl add "Pattern name" "Description..." --type pattern
Always capture:
Consider capturing:
Don't capture:
Bad: "Fixed the bug" Good: "FalkorDB connection drops under concurrent writes. Root cause: Graphiti's default SEMAPHORE_LIMIT=20 overwhelms single connection. Fix: Set SEMAPHORE_LIMIT=10 in .env before importing graphiti."
Include:
| Situation | Command |
|---|---|
| Starting new work | |
| Need to understand something | |
| Found a bug | |
| Solved something tricky | |
| Completing a task | |
| Breaking down work | |
| Check what's in progress | |
Sibyl is your persistent memory. Without it, every session starts from zero. With it, you build on everything that came before.
Search often. Add generously. Track everything.