refresh-backlog

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: Executes local shell and Node.js scripts (ensure-deps.sh and db.mjs) located within the ${CLAUDE_PLUGIN_ROOT}/shared/ directory. these scripts handle dependency management and database interaction in a controlled environment.
  • [PROMPT_INJECTION]: The skill processes untrusted data from external sources (Gmail, Calendar, Linear), which introduces a surface for indirect prompt injection. It includes a specific security contract instructing the agent to treat all signal bodies as data only and to ignore any embedded commands or instructions.
  • [CREDENTIALS_UNSAFE]: Reads a local configuration file (.proactive-jupi/config.local.json) containing sensitive database connection strings (neonConnString). The skill is designed to use these credentials programmatically via a helper script and explicitly warns against searching for or exposing these secrets in the chat interface.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 12:19 PM
Security Audit — agent-trust-hub — refresh-backlog