prompt

Overview

Captures every prompt (and optional role/kind) as a memory episode in Graphiti MCP for a repo/user. This enables agents to recall recent context and maintain a searchable history of prompts.

How This Skill Works

Prompts are captured automatically when triggers like load prompt or capture prompt fire, or via per-prompt hook integrations. The capture uses the GRAPHITI_ENDPOINT from the .lisa/.env (with a fallback to http://localhost:8010/mcp/), and derives a group ID from the project folder path. A fingerprint tag prevents duplicate prompts unless --force is provided, and the operation outputs a JSON status/action.

When to Use It

• When you want every prompt stored automatically for a repo/user to improve recall.
• When prompts vary across roles and kinds and you want that metadata preserved.
• When you need to audit or review past prompts for debugging or compliance.
• When recalling recent context is critical for agent decision making across sessions.
• When integrating with triggers like load prompt or per-prompt hooks to ensure automatic capture.

Quick Start

1. Step 1: Run lisa-prompt add --text "Your prompt here" [--role user|assistant] [--kind Direction|Decision|Requirement|Observation] [--force]
2. Step 2: Ensure GRAPHITI_ENDPOINT is set in .lisa/.env (or rely on the default http://localhost:8010/mcp/). The group ID is derived from the project folder path.
3. Step 3: Check the CLI output for a JSON status/action and verify a fingerprint tag was applied.

Best Practices

• Enable fingerprinting to avoid duplicates; use --force to overwrite when necessary.
• Set GRAPHITI_ENDPOINT in the .lisa/.env file or rely on the default http://localhost:8010/mcp/.
• Use --role and --kind to structure metadata (Direction, Decision, Requirement, Observation).
• Let the group ID be derived from the project folder path for consistent organization.
• Test with a small batch of prompts before enabling widespread automatic capture.

Example Use Cases

• User: Outline a plan to migrate service X to Graphiti-based workflow.
• User: What criteria should we use to evaluate feature flag rollout strategy?
• User: Define acceptance criteria for bug fix in module X.
• User: Capture notes on edge-case handling for input Y during validation.
• User: What questions should be asked to clarify the user story Z?

Frequently Asked Questions