cognee-sync

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill runs a local script located at ${CLAUDE_PLUGIN_ROOT}/scripts/sync-session-to-graph.py to process and persist session data into a permanent knowledge graph.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes session cache entries and session text, which are untrusted data sources. This creates a surface for indirect prompt injection where malicious content in the session could influence the knowledge graph metadata or embeddings.
  • Ingestion points: Session cache entries and session Q&A text processed during synchronization (SKILL.md).
  • Boundary markers: None explicitly defined in the skill instructions to delimit untrusted session content from the synchronization logic.
  • Capability inventory: Execution of a local Python script to update the knowledge graph, enrich triplets, and update vector search indices.
  • Sanitization: No specific sanitization, filtering, or validation of session content is described in the provided documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 05:49 AM
Security Audit — agent-trust-hub — cognee-sync