knowledge-graph

Pass

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSMETADATA_POISONING
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and stores observations about external entities in a persistent knowledge graph, which are later retrieved and injected into the agent's context, creating a potential vector for indirect prompt injection.\n
  • Ingestion points: Observations are ingested via the mcp__memory__add_observations tool and the add_entity function in the local JSON implementation.\n
  • Boundary markers: The instructions do not specify the use of delimiters or boundary markers to isolate stored observations from the system prompt.\n
  • Capability inventory: The skill allows use of tools including Bash, Write, and Edit.\n
  • Sanitization: There is no evidence of content sanitization or validation performed on observations before they are stored or retrieved.\n- [EXTERNAL_DOWNLOADS]: The skill setup involves downloading and executing the Model Context Protocol memory server from a public registry.\n
  • Evidence: Configuration instructions include npx -y @modelcontextprotocol/server-memory in SKILL.md.\n- [METADATA_POISONING]: The skill includes metadata fields that make self-verifying safety claims which cannot be externally validated.\n
  • Evidence: The YAML frontmatter includes verified: true and trust_score: 100 fields.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 12, 2026, 12:45 PM