Graphiti

PassAudited by ClawScan on May 1, 2026.

Overview

The skill is a coherent Graphiti API wrapper with no artifact-backed malicious behavior, but it can write persistent graph memory and has a URL-configuration inconsistency to notice.

Install this only if you want the agent to query and add data to your Graphiti knowledge graph. Treat added episodes as persistent memory, avoid storing secrets, and verify the Graphiti endpoint with Clawdbot config rather than relying on the documented GRAPHITI_URL environment fallback.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

The agent may add or change remembered information in the connected Graphiti service.

Why it was flagged

The skill uses shell/curl instructions to perform a Graphiti API write. This is aligned with the stated purpose, but users should notice that the agent can create new knowledge-graph entries.

Skill content
curl -s -X POST "$GRAPHITI_URL/messages" ... -d '{"name": "EPISODE_NAME", "content": "EPISODE_CONTENT"}'
Recommendation

Use the add operation only for content you intentionally want stored, and review important additions before relying on them later.

What this means

Sensitive, incorrect, or adversarial content stored as a memory could be retrieved and reused in future tasks.

Why it was flagged

The skill intentionally writes persistent memory to a knowledge graph. Persistent memories can later influence search results and agent context.

Skill content
Add a new episode/memory to the knowledge graph.
Recommendation

Avoid storing secrets or untrusted instructions as memories, and periodically review or remove outdated Graphiti entries.

What this means

If a user sets GRAPHITI_URL expecting a specific service, the skill may still use Clawdbot config or localhost instead.

Why it was flagged

The script overwrites GRAPHITI_URL from Clawdbot config before falling back to localhost, despite documentation saying the environment variable is supported. This can make endpoint selection differ from user expectations.

Skill content
GRAPHITI_URL=$(clawdbot config get skills.graphiti.baseUrl 2>/dev/null || echo "") ... GRAPHITI_URL=${GRAPHITI_URL:-"http://localhost:8001"}
Recommendation

Prefer setting `clawdbot config set skills.graphiti.baseUrl ...` and verify the endpoint before adding memories.