cekura-generate-scenarios

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted production call logs to generate evaluator scenarios.
  • Ingestion points: Transcript data and metadata are ingested from mcp__cekura__call_logs_retrieve tool results.
  • Boundary markers: Absent; the instructions do not provide delimiters or warnings for the agent to ignore embedded instructions within call transcripts.
  • Capability inventory: The skill has access to write files, execute bash commands, and create or update scenarios and test profiles through MCP tools.
  • Sanitization: Absent; there is no specified mechanism for sanitizing or escaping transcript content before it is interpolated into prompts or passed to generation endpoints.
  • [DATA_EXFILTRATION]: The skill extracts sensitive PII, including names, dates of birth, ZIP codes, and medications, from call logs to populate test profiles. While this is a primary function, it involves handling sensitive identity data across platform boundaries.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to save scenario reports locally as markdown files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:13 PM
Security Audit — agent-trust-hub — cekura-generate-scenarios