cognee-search
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves data from a remote or local Cognee server and incorporates it into the agent's context. This creates an attack surface where instructions embedded in indexed documents or memory could attempt to influence the agent's behavior.
- Ingestion points:
SKILL.md(POST requests to recall API andcognee-clifallback) - Boundary markers: Not specified in the skill instructions; retrieved data is directly provided as context
- Capability inventory: Shell command execution (
cognee-search.sh,cognee-cli), network communication viacurlto external or local endpoints - Sanitization: No explicit sanitization or filtering of the retrieved content is described
- [COMMAND_EXECUTION]: The skill executes local shell scripts (
cognee-search.sh) and a CLI utility (cognee-cli) to process searches and handle server fallbacks.
Audit Metadata