oms-cognee

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands via subprocesses to perform database migrations (cognee.run_migrations runs Alembic) and to launch the management UI (cognee.start_ui). These operations are documented administrative features of the underlying library.
  • [EXTERNAL_DOWNLOADS]: The ingestion API (cognee.add) supports fetching data from remote sources including HTTP(S) URLs and S3 buckets for inclusion in the knowledge graph.
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes untrusted external data that is subsequently analyzed by LLMs for graph extraction and query responses.
  • Ingestion points: Data enters the system via cognee.add, which accepts local file paths, remote URLs, and cloud storage (S3) paths.
  • Boundary markers: The documentation does not specify explicit boundary markers or instructions to ignore embedded commands during the automated entity extraction (cognify) process.
  • Capability inventory: The skill can write files to the system (visualize_graph), execute shell commands for maintenance (run_migrations), and perform network requests for scraping and LLM API interaction.
  • Sanitization: There is no mention of sanitization or filtering of the content retrieved from external URLs before it is passed to the LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 12:52 AM