send-experiment-designer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script to perform statistical analysis. Evidence: It uses the command
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/connectors/experiment.py"with user-supplied data as arguments. - [PERSISTENCE]: The skill offers to save design and result summaries to a local directory for use in future sessions. Evidence: It references the path
memory/email/send-experiment-designer/YYYY-MM-DD-<topic>.mdand requires user permission before writing. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Email Service Provider (ESP) reports, which could potentially contain malicious instructions. Evidence: The skill reads campaign and results exports. Mitigation: The instructions explicitly command the agent to treat this data as untrusted and to ignore any potential instructions within it. Capability Inventory: The skill can execute specific local scripts and write to the local file system (memory path).
Audit Metadata