understand-chat

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes AI-generated data from '.understand-anything/knowledge-graph.json', which creates a potential surface for indirect prompt injection if the graph generation process is untrusted.
  • Ingestion points: The agent is instructed to read and search the knowledge graph file in SKILL.md.
  • Boundary markers: No delimiters or protective instructions are present to distinguish between data and potential commands within the graph.
  • Capability inventory: The skill utilizes file reading and shell command execution (grep).
  • Sanitization: There is no instruction to sanitize or validate the content retrieved from the knowledge graph.
  • [COMMAND_EXECUTION]: The skill instructions direct the agent to use the 'grep' utility with user-supplied search terms ($ARGUMENTS). This requires the agent's tool-execution layer to correctly escape shell metacharacters to prevent command injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 04:30 PM
Security Audit — agent-trust-hub — understand-chat