cognee-docker

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external data into an AI knowledge graph, creating a surface for indirect prompt injection. * Ingestion points: Data is uploaded via the '/api/v1/remember' endpoint using the curl command found in SKILL.md. * Boundary markers: Absent; no delimiters or instructions to ignore embedded content are provided. * Capability inventory: The skill utilizes docker compose for container orchestration and curl for network interactions with the local API. * Sanitization: Absent; the skill does not perform content validation or escaping on ingested documents.
  • [CREDENTIALS_UNSAFE]: The Neo4j database profile uses a hardcoded default password ('pleaseletmein') for local development access.
  • [EXTERNAL_DOWNLOADS]: The skill downloads the 'cognee/cognee:main' image from Docker Hub, which is the official repository for the vendor.
  • [COMMAND_EXECUTION]: Shell commands such as 'docker compose up' and 'curl' are used to manage the service and interact with its API.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 06:30 PM